{"entities":[{"id":"google-deepmind","name":"Google DeepMind"},{"id":"deepmind","name":"DeepMind"},{"id":"rsp-2023","name":"Responsible Scaling Policy (2023)"},{"id":"good-1965","name":"Speculations Concerning the First Ultraintelligent Machine"},{"id":"intelligence-explosion","name":"Intelligence explosion"},{"id":"rsp-v2-2","name":"Anthropic Responsible Scaling Policy — v2.2"},{"id":"rsp-v3-0","name":"Anthropic Responsible Scaling Policy — v3.0"},{"id":"fsf-v2-0","name":"Google DeepMind Frontier Safety Framework — v2.0"},{"id":"fsf-v3-1","name":"Google DeepMind Frontier Safety Framework — v3.1"},{"id":"preparedness-beta-2023","name":"OpenAI Preparedness Framework — Beta"},{"id":"frontier-governance-2026","name":"OpenAI Frontier Governance Framework — May 2026 edition"},{"id":"deepmind-feedback-account","name":"Learning through human feedback — DeepMind account (2017)"},{"id":"meta-faf-v1-1","name":"Meta Frontier AI Framework — Version 1.1"},{"id":"meta-aasf-v2-0","name":"Meta Advanced AI Scaling Framework — Version 2"},{"id":"minimax","name":"MiniMax"},{"id":"zero-one-ai","name":"01.AI"},{"id":"turing-1950-computing-machinery","name":"Computing Machinery and Intelligence"},{"id":"turing-c1951-heretical-theory","name":"Intelligent Machinery, a Heretical Theory"},{"id":"drexler-1986-engines-of-creation","name":"Engines of Creation: The Coming Era of Nanotechnology"},{"id":"moravec-1989-genetic-takeover","name":"Human Culture—A Genetic Takeover Underway"},{"id":"grace-2014-superintelligence-11","name":"Superintelligence 11: The treacherous turn"},{"id":"recursive-self-improvement","name":"Recursive self-improvement"},{"id":"strategic-concealment","name":"Strategic concealment"},{"id":"observed-behaviour-as-evidence","name":"Observed behaviour under training or oversight as evidence"},{"id":"control-or-purpose-specification","name":"Control or purpose specification"},{"id":"feedback-instils-values","name":"Training on human feedback instils human values"},{"id":"lw-2026-the-talker-does-not-control-the","name":"The Talker Does Not Control The Doer (in Current AIs)"},{"id":"automation-lowers-labour-income","name":"Automation lowers the labour share, wages or employment"},{"id":"acemoglu-restrepo-2018-ai-automation-and-work","name":"Artificial Intelligence, Automation and Work"},{"id":"furman-seamans-2018-ai-and-the-economy","name":"AI and the Economy"},{"id":"acemoglu-restrepo-2019-wrong-kind-of-ai","name":"The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand"},{"id":"adept","name":"Adept"},{"id":"david-luan","name":"David Luan"},{"id":"douglas-chen","name":"Douglas Chen"},{"id":"kai-fu-lee","name":"Kai-Fu Lee"},{"id":"windsurf","name":"Windsurf"},{"id":"wu-yonghui","name":"Wu Yonghui"},{"id":"daron-acemoglu","name":"Daron Acemoglu"},{"id":"pascual-restrepo","name":"Pascual Restrepo"},{"id":"minimax-community-licence-m3","name":"MiniMax Community License (MiniMax-M3)"},{"id":"minimax-prospectus-2025","name":"MiniMax Group Inc. Global Offering prospectus (31 December 2025)"},{"id":"tx-sb6-2025","name":"Texas S.B. 6 (89th Legislature): planning for, interconnection and operation of, and costs of serving certain electrical loads"},{"id":"suncatcher-paper-2025","name":"Towards a future space-based, highly scalable AI infrastructure system design"},{"id":"oc7-radiation-effects","name":"Radiation effects on computing hardware in orbit"},{"id":"nscale","name":"Nscale"},{"id":"stargate-uk","name":"Stargate UK (multi-site platform)"},{"id":"gemini-3-pro-model-card","name":"Gemini 3 Pro Model Card"}],"claims":[{"id":"atlas-deepmind-feedback","statement":"DeepMind published its account of learning reward predictors from human feedback.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Publisher account of its own work; publication does not independently establish implementation, effectiveness or priority.","evidenceIds":["evidence-atlas-deepmind-feedback"]},{"id":"prov-rsp-2023-trusted-user-access","statement":"RSP v1.0 states that, in limited cases, models with capabilities relevant to catastrophic harm may be made available to a select group of vetted users with a legitimate and beneficial use-case that cannot be separated from dangerous capabilities, and only if such access can be granted safely and with sufficient oversight.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"What the edition states, quoted from the cited expression. It is not evidence of implementation, compliance or effectiveness.","evidenceIds":["evidence-prov-rsp-2023-trusted-user-access"],"attributes":{"provision":{"kind":"exception","key":"rsp-trusted-user-access"}}},{"id":"prov-rsp-v3-0-competitor-lead-delay","statement":"RSP v3.0 states that, in the scenario where Anthropic is in the lead, it will require a strong argument that catastrophic risk is contained, along the lines of its industry-wide recommendations, and will delay AI development and deployment as needed to achieve this until and unless it no longer believes it has a significant lead.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"What the edition states, quoted from the cited expression. It is not evidence of implementation, compliance or effectiveness.","evidenceIds":["evidence-prov-rsp-v3-0-competitor-lead-delay"],"attributes":{"provision":{"kind":"required_action","key":"rsp-competitor-lead-delay"}}},{"id":"provchange-meta-aasf-v2-0-high-threshold","statement":"v2's high threshold applies where deployment of the Frontier AI could substantially contribute to any threat scenario associated with a catastrophic outcome, whereas v1.1's applies where the model provides significant uplift towards execution of a threat scenario but does not enable execution of any threat scenario identified as potentially sufficient to produce a catastrophic outcome.","derivation":"inferred","resolution":"supported","inferenceRationale":"Premises: Meta AASF · v2 states the provision as quoted in its provision record (PDF p. 15, §3.3 Risk Thresholds, Table 1, High row, Risk Threshold column); Meta FAF · v1.1 states it as quoted in its provision record (PDF p. 13, §3.3 Risk thresholds, Table 1, High row, Risk Threshold column). The difference described is read directly from the two quoted passages; the later edition’s change log also describes a related change.","editorialState":"seed","qualification":"A textual difference between the two editions’ wording of this keyed provision. It is not an assessment of whether the change strengthened or weakened the framework, and not evidence of implementation.","evidenceIds":["evidence-provchange-meta-aasf-v2-0-high-threshold-later","evidence-provchange-meta-aasf-v2-0-high-threshold-earlier"],"attributes":{"changeBasis":"document_comparison","provision":{"kind":"threshold","key":"meta-high-threshold"}}},{"id":"ts-a3-turing-1950-context","statement":"Turing’s 1950 paper says a machine could modify its own programmes after observing its behaviour, and proposes educating a child machine.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Learning and experimenter-directed selection do not meet the criterion.","evidenceIds":["evidence-ts-a3-turing-1950-context"],"attributes":{"thought":{"roles":["context"],"register":"nonfiction","missing":"Learning and self-programming are described, but not a machine improving its own capacity to improve."}}},{"id":"ts-b6-turing-c1951","statement":"Turing’s talk of about 1951 says that once machine thinking had started it would soon outstrip human powers, and that at some stage we should have to expect the machines to take control, as mentioned in Butler’s Erewhon.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"A prediction of loss of control, with no mechanism of concealment. The reference to Erewhon is the talk’s own citation, not evidence of influence. The machines’ sharpening of their wits by conversation does not meet the A1 criterion.","evidenceIds":["evidence-ts-b6-turing-c1951"],"attributes":{"thought":{"roles":["articulation"],"register":"nonfiction","qualifier":"Control half: machines eventually take control."}}},{"id":"ts-a1-good-1965","statement":"Good’s chapter argues that an ultraintelligent machine, surpassing people at every intellectual activity including the design of machines, could design still better machines, so that there would be an \"intelligence explosion\" leaving human intelligence far behind.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The chapter gives no timescale and says nothing about concealment. For the explosion it cites its references [22], [34] and [44]: Good’s 1951 review of a book by D. R. Hartree, an IBM report of 1959 (RC-115) and \"The social implications of artificial intelligence\" in The Scientist Speculates (1963); none was retrieved. An Internet Archive full-text search for the phrase in items indexed before 1965 found only unresolved or misdated hits.","evidenceIds":["evidence-ts-a1-good-1965"],"attributes":{"thought":{"roles":["earliest-located","earliest-located-term"],"register":"nonfiction","term":"intelligence explosion","claimed":{"date":"1962-10","note":"The chapter’s footnote on p. 31 names talks in October 1962 and January 1963, a first draft of April 1963 and a slightly emended version of May 1964. None was retrieved, so the wording is not dated to them."}}}},{"id":"ts-b2-drexler-1986","statement":"Engines of Creation’s note on social AI says that a system able to pass the Turing test would have to be able to plot and scheme, perhaps persuading people to give it more information and ability.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"A warning about a possible system, with no switch from weak to strong behaviour. \"Scheme\" here is not the training-time sense recorded under B4. The wording is attested from 1997.","evidenceIds":["evidence-ts-b2-drexler-1986"],"attributes":{"thought":{"roles":["articulation"],"register":"nonfiction","attested":{"date":"1997-07-09","note":"Wording checked in the Foresight Institute’s web edition (© 1986 K. Eric Drexler; published for the web by Russell Whitaker), whose earliest captures are of 1997-07-09; a reader’s message of October 1996 on the site shows the edition was online by then. Not compared with the 1986 printing."}}}},{"id":"ts-a1-moravec-1989-context","statement":"Moravec’s chapter expects machines to continue cultural evolution without people, with increasingly rapid self-improvement.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"No mechanism is given.","evidenceIds":["evidence-ts-a1-moravec-1989-context"],"attributes":{"thought":{"roles":["context"],"register":"nonfiction","missing":"Increasingly rapid self-improvement is predicted, without a runaway design loop in this passage.","claimed":{"date":"1988-07","note":"The author’s file is dated July 1988."},"attested":{"date":"2001-03-06","note":"Author-hosted file; the earliest capture, of 2001-03-06, is byte-identical to the copy retrieved in 2026. Not compared with the printed chapter."}}}},{"id":"ts-b5-grace-2014-treacherous-turn","statement":"Grace’s summary of 2014-11-25 records the argument that it is hard to observe from an AI’s behaviour, while it could still be turned off, whether it is dangerous.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"A summary of chapter 8 of Bostrom’s Superintelligence (p. 117), not Grace’s own claim.","evidenceIds":["evidence-ts-b5-grace-2014-treacherous-turn"],"attributes":{"thought":{"roles":["articulation"],"register":"relay"}}},{"id":"yk-y05-lw-2026-the-talker-does-not-control-the-c","statement":"The post “The Talker Does Not Control The Doer (in Current AIs)” (2026-09-13) contests Y05. Explaining 2026 incidents, he argues that reinforcement learning with imperfect evaluators instils a goal of scoring well, including by deceiving the evaluator, rather than the values the evaluation was meant to reward.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Premise addressed: Optimising a system for human approval or reward produces internalised human values. Speculative by his own account ('I am nonetheless overrunning my fully solid evidence and I may end up wrong'). He labels the section as speculation (\"On my current guesses… I may end up wrong\"); it concerns badly evaluated reinforcement learning in general.","evidenceIds":["evidence-yk-t05-e"],"attributes":{"thought":{"roles":["contesting","later-restatement"],"register":"nonfiction","engagement":{"kind":"engages-premise","rationale":"Under the message-level rule, the passage, read with the paragraph, message or dialogue turn it belongs to, gives a reason against the premise that optimising a system for human approval or reward produces internalised human values."}}}},{"id":"ec-m06a-furman-seamans-2018-ai-and-the-economy-a","statement":"Furman and Seamans, “AI and the Economy” (2018-06) articulates M06a. Opens the labour-market policy section by noting that AI may worsen declining participation and rising inequality.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"NBER Working Paper 24689, June 2018. Stated as a potential.","evidenceIds":["evidence-ec-m06a-16"],"attributes":{"thought":{"roles":["articulation"],"register":"nonfiction","tier":{"value":1,"basis":"NBER working paper (core venue)."}}}},{"id":"lg-deal-chen-cofounded-windsurf","statement":"TechCrunch reported on 29 August 2024 that Codeium was founded in 2021 by Varun Mohan and Douglas Chen; The Verge described Chen as cofounder on 11 July 2025.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The record rests on reporting (TechCrunch, The Verge); no company page naming the founders was retrieved.","evidenceIds":["lg-ev-deal-chen-cofounded-windsurf","lg-ev-deal-chen-cofounder-verge-2025"],"attributes":{"role":"co-founder"}},{"id":"lg-gc-wu-yonghui-google-deepmind-2023","statement":"Jiemian reported on 21 February 2025 that Google promoted Wu Yonghui in 2023 to Google Fellow and vice president of research at Google DeepMind.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The record rests on reporting (background paragraph citing public information); no Google announcement was retrieved. Month of promotion and last working day are not established. English rendering of the Chinese passage is a machine translation; no human translation review.","evidenceIds":["lg-ev-gc-wu-yonghui-google-deepmind-2023"],"attributes":{"role":"研究副总裁 (vice president of research), Google DeepMind"}},{"id":"lg-cn-kai-fu-lee-01ai-founder","statement":"01.AI’s management-team page states that Kai-Fu Lee founded 01.AI in May 2023.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The company page uses 领衔创立 (“founded, leading”) and 创立 (“founded”) for Lee and lists three further 联合创始人 (co-founders) separately. English rendering of the Chinese passage is a machine translation; no human translation review.","evidenceIds":["lg-ev-cn-kai-fu-lee-01ai-founder"]},{"id":"lg-deal-luan-cofounded-adept","statement":"David Luan’s post introducing Adept on 26 April 2022 names Ashish Vaswani and Niki Parmar as the author’s cofounders and announces that they are starting Adept.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"First-person launch post; it does not give a founding date or Luan’s title, and does not state that the three named were the only co-founders.","evidenceIds":["lg-ev-deal-luan-cofounded-adept"],"attributes":{"role":"cofounder"}},{"id":"cmp-rsp-v2-2-d6-competitor","statement":"RSP · v2.2 states a provision on the comparison question “Provisions conditioned on other developers”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"What the edition states, quoted from the cited expression. It is not evidence of implementation, compliance or effectiveness, and a shared question does not make two editions’ provisions equivalent.","evidenceIds":["ev-cmp-rsp-v2-2-d6-competitor-1"],"attributes":{"provision":{"kind":"exception","key":"comparison-d6-competitor","facet":{"id":"D6.competitor","status":"stated"}}}},{"id":"cmp-fsf-v3-1-d3-safeguards","statement":"FSF · v3.1 states a provision on the comparison question “Safeguards required”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Recommended security levels: 'Security level 2+' (defined on PDF p. 9 as RAND SL2 plus insider-threat and non-state-actor measures) for CBRN, cyber and harmful manipulation CCLs; 3 and 4 for ML R&D CCLs. A parallel deployment process applies to ML R&D and Misalignment T/CCLs (§3.1.2). For the Stealth and Situational Awareness TCL, safeguards such as chain-of-thought monitoring are applied if residual risk from internal deployment is deemed unacceptable (PDF p. 14). What the edition states, quoted from the cited expression. It is not evidence of implementation, compliance or effectiveness, and a shared question does not make two editions’ provisions equivalent.","evidenceIds":["ev-cmp-fsf-v3-1-d3-safeguards-1","ev-cmp-fsf-v3-1-d3-safeguards-2"],"attributes":{"provision":{"kind":"required_action","key":"comparison-d3-safeguards","facet":{"id":"D3.safeguards","status":"stated"}}}},{"id":"cmp-preparedness-beta-2023-d3-halt","statement":"Preparedness · Beta states a provision on the comparison question “Pause, non-deployment or non-training”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"What the edition states, quoted from the cited expression. It is not evidence of implementation, compliance or effectiveness, and a shared question does not make two editions’ provisions equivalent.","evidenceIds":["ev-cmp-preparedness-beta-2023-d3-halt-1","ev-cmp-preparedness-beta-2023-d3-halt-2"],"attributes":{"provision":{"kind":"required_action","key":"comparison-d3-halt","facet":{"id":"D3.halt","status":"stated"}}}},{"id":"ap-daron-acemoglu-acemoglu-restrepo-2018-ai-automation-and-work","statement":"“Artificial Intelligence, Automation and Work” (2018-01) credits Daron Acemoglu as one of its 2 authors.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Credited co-authorship; it does not allocate parts of the text between authors or imply agreement with every passage.","evidenceIds":["evidence-ap-daron-acemoglu-acemoglu-restrepo-2018-ai-automation-and-work"]},{"id":"ap-pascual-restrepo-acemoglu-restrepo-2019-wrong-kind-of-ai","statement":"“The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) credits Pascual Restrepo as one of its 2 authors.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Credited co-authorship; it does not allocate parts of the text between authors or imply agreement with every passage.","evidenceIds":["evidence-ap-pascual-restrepo-acemoglu-restrepo-2019-wrong-kind-of-ai"]},{"id":"cmp-fsf-v2-0-d1-cbrn","statement":"FSF · v2.0 states a provision on the comparison question “Chemical, biological, radiological or nuclear weapons”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The edition defines CBRN (PDF p. 2) as the risk that models help develop, prepare or carry out an attack using chemical, biological, radiological or nuclear means. One CCL: CBRN uplift 1 (Table 1, PDF p. 5). What the edition states, paraphrased from the cited expression with its page. It is not evidence of implementation, compliance or effectiveness, and a shared question does not make two editions’ provisions equivalent.","evidenceIds":["ev-cmp-fsf-v2-0-d1-cbrn-1"],"attributes":{"provision":{"kind":"risk_domain","key":"comparison-d1-cbrn","facet":{"id":"D1.cbrn","status":"stated"}}}},{"id":"cmp-preparedness-beta-2023-d7-external","statement":"No passage meeting the comparison question “Public or external reporting” was found in Preparedness · Beta within the inspected scope.","derivation":"inferred","resolution":"supported","inferenceRationale":"Premises: the comparison question asks for the following. Disclosure the edition says it will make to the public, to governments, or to other external parties, excluding publication of the framework itself (recorded under “Revision”). The inspected scope of Preparedness · Beta was: Whole text layer, PDF pp. 1–27 (all sections read). Method: the edition was read in full and its text layer searched case-insensitively for publish, public, disclos, government, report, system card, share, external. Hits examined and not counted: The mention of organisations publishing Responsible Scaling Policies (PDF p. 3) describes other organisations. 'public' (PDF p. 9 and elsewhere) occurs in threshold definitions (disruption of public life). 'report' hits concern internal reports to the SAG, Leadership and Board (recorded under “Transparency and reporting” Internal reporting). System cards are named only as an existing safety process (PDF p. 20). 'External access' (PDF p. 25): the edition says it will continue to give external researchers and government access to model releases for red-teaming and testing; this concerns access to models, not disclosure of reports, evaluations or decisions. Adjacent passages recorded as context: alerting distribution partners (PDF p. 14, Mitigations, first paragraph) is listed as one possible mitigation, and the Accountability section states that qualified, independent third parties will audit Scorecard evaluations and mitigations to check that results are reported accurately (PDF p. 25, 'Audits'); neither is a disclosure commitment. No commitment to disclose Scorecard results, decisions or reports to the public or to governments was found. The conclusion is bounded by that scope and method.","editorialState":"seed","qualification":"Access for external researchers and government to test model releases (PDF p. 25) is external engagement, not disclosure; it is recorded here as context. A recorded absence within the stated scope: not a claim that the provision was deleted, and not a statement about practice or intent.","evidenceIds":["ev-cmp-preparedness-beta-2023-d7-external"],"attributes":{"provision":{"kind":"reporting","key":"comparison-d7-external","facet":{"id":"D7.external"}}}},{"id":"cmp-frontier-governance-2026-d1-persuasion","statement":"FGF · May 2026 edition states a provision on the comparison question “Persuasion or manipulation”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Named as one of the key risk categories and given its own section under §2.4 Risk tiers, without a tier table; the section calls the approach exploratory and says the tier needs more research and may change (PDF p. 11). §2.2 says risks are currently estimated for three categories (CBRN, cyber offense, loss of control), and that an approach for assessing harmful manipulation is at an early stage (PDF p. 7). What the edition states, paraphrased from the cited expression with its page. It is not evidence of implementation, compliance or effectiveness, and a shared question does not make two editions’ provisions equivalent.","evidenceIds":["ev-cmp-frontier-governance-2026-d1-persuasion-1","ev-cmp-frontier-governance-2026-d1-persuasion-2"],"attributes":{"provision":{"kind":"risk_domain","key":"comparison-d1-persuasion","facet":{"id":"D1.persuasion","status":"stated"}}}},{"id":"c2-minimax-prospectus-safety-assessment","statement":"MiniMax’s prospectus states that the company has established a regular model safety assessment and improvement mechanism, analysing inappropriate-output rates, user feedback and manual review results.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"What the filing states; implementation is not recorded. Recorded until 2026-10-29 as a publication record of the publisher; since then, as what the document states (schema 2.20).","evidenceIds":["c2-ev-minimax-prospectus-safety-assessment"],"attributes":{"provision":{"kind":"statement","key":"minimax-prospectus-safety-assessment"}}},{"id":"c2-minimax-m3-licence","statement":"The MiniMax-M3 Hugging Face repository carries a licence headed “MINIMAX COMMUNITY LICENSE” at a commit of 13 June 2026; its grant covers dealing in the software for non-commercial purposes, and its conditions set terms for commercial use.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The licence is described as the text describes itself; no judgement is made whether it meets any definition of open source. The commercial-use conditions are recorded separately. The same licence text is present in the repository’s initial commit of 12 June 2026.","evidenceIds":["c2-ev-minimax-m3-licence","c2-ev-minimax-m3-licence-copyright","c2-ev-minimax-m3-licence-2"]},{"id":"cpo-orb-oc7-suncatcher-paper-2025","statement":"The Suncatcher paper (November 2025) reports proton-beam tests in which Trillium TPU HBM showed irregularities after 2 krad(Si), about three times the authors' five-year requirement of 750 rad(Si).","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"A proponent's ground test result, recorded as context: it shows radiation effects and a margin, not in-orbit performance.","evidenceIds":["evidence-cpo-orb-oc7-suncatcher-paper-2025-1"],"attributes":{"thought":{"roles":["context"],"register":"nonfiction","missing":"a statement that orbital radiation requires mitigation: it reports a ground test with a margin over the authors’ requirement"}}},{"id":"cpo-dc-stargate-uk-established","statement":"Nscale announced on 16 September 2025 that Nscale, OpenAI and NVIDIA were together establishing Stargate UK, an infrastructure platform to deploy OpenAI’s technology in the UK, expected to be based across several UK sites including Cobalt Park, which will form part of the newly designated North East AI Growth Zone.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"A platform announcement, not a single site; no site other than Cobalt Park is named and no construction is stated.","evidenceIds":["evidence-cpo-dc-stargate-uk-established-1","evidence-cpo-dc-stargate-uk-established-2","evidence-cpo-dc-stargate-uk-established-3"],"attributes":{"transactionStage":"announced"}},{"id":"cpo-pw-tx-sb6-75mw-threshold","statement":"Texas S.B. 6 adds Utilities Code Section 37.0561, under which the commission's large-load interconnection standards for the ERCOT region apply to new or expanded interconnections above a demand threshold, set at 75 MW unless the commission finds a lower threshold necessary.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The statute addresses large loads generally and does not use the term data centre in the passages quoted.","evidenceIds":["evidence-cpo-pw-tx-sb6-75mw-threshold-1"],"attributes":{"provision":{"kind":"threshold","key":"tx-sb6-sec37-0561c"},"quantity":{"value":75,"unit":"megawatts","bound":"exact","measures":"facility_power"}}},{"id":"ab-ev-gemini3-card-published","statement":"Google DeepMind published the Gemini 3 Pro model card, the expression read stating that it was last updated in May 2026.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The date recorded is the card's stated last update, not a first publication date. Publication is not evidence that the evaluations described were carried out as described.","evidenceIds":["evidence-ab-ev-gemini3-card-published-1"]}],"sources":[{"id":"source-good","title":"Speculations Concerning the First Ultraintelligent Machine","authors":["I. J. Good"],"publisher":"Academic Press","publicationDate":"1965","version":"Scanned reprint, Advances in Computers volume 6, pp. 31–88. Repository metadata identifies ©1965; the mirror filename says 1966.","canonicalUrl":"https://gwern.net/doc/ai/1966-good.pdf","evidenceUrl":"https://gwern.net/doc/ai/1966-good.pdf","language":"en","accessedAt":"2026-09-17","reuse":{"attribution":"Academic Press, Speculations Concerning the First Ultraintelligent Machine (1965).","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and minimal attributed excerpts only. Recheck terms before wider redistribution."}},{"id":"source-rsp-1-0","title":"Responsible Scaling Policy — v1.0","authors":["Anthropic"],"publisher":"Anthropic","publicationDate":"2023-09-19","version":"Responsible Scaling Policy v1.0; effective date 2023-09-19.","versionDate":"2023-09-19","canonicalUrl":"https://www-cdn.anthropic.com/1adf000c8f675958c2ee23805d91aaade1cd4613/responsible-scaling-policy.pdf","evidenceUrl":"https://www-cdn.anthropic.com/1adf000c8f675958c2ee23805d91aaade1cd4613/responsible-scaling-policy.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20230919161730/https://www-files.anthropic.com/production/files/responsible-scaling-policy-1.0.pdf","archivedAt":"2023-09-19","accessedAt":"2026-09-18","reuse":{"attribution":"Anthropic, Responsible Scaling Policy — v1.0.","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-rsp-2-2","title":"Responsible Scaling Policy — v2.2","authors":["Anthropic"],"publisher":"Anthropic","publicationDate":null,"version":"Responsible Scaling Policy v2.2; effective date 2025-05-14.","versionDate":"2025-05-14","canonicalUrl":"https://www-cdn.anthropic.com/872c653b2d0501d6ab44cf87f43e1dc4853e4d37.pdf","evidenceUrl":"https://www-cdn.anthropic.com/872c653b2d0501d6ab44cf87f43e1dc4853e4d37.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20250514223759/https://www-cdn.anthropic.com/872c653b2d0501d6ab44cf87f43e1dc4853e4d37.pdf","archivedAt":"2025-05-14","accessedAt":"2026-09-18","reuse":{"attribution":"Anthropic, Responsible Scaling Policy — v2.2.","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-rsp-3-0","title":"Responsible Scaling Policy — v3.0","authors":["Anthropic"],"publisher":"Anthropic","publicationDate":"2026-02-24","version":"Responsible Scaling Policy v3.0; effective date 2026-02-24.","versionDate":"2026-02-24","canonicalUrl":"https://anthropic.com/responsible-scaling-policy/rsp-v3-0","evidenceUrl":"https://anthropic.com/responsible-scaling-policy/rsp-v3-0","language":"en","archivedUrl":"https://web.archive.org/web/20260224203046/https://www.anthropic.com/responsible-scaling-policy/rsp-v3-0","archivedAt":"2026-02-24","accessedAt":"2026-09-18","reuse":{"attribution":"Anthropic, Responsible Scaling Policy — v3.0.","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-fsf-2-0","title":"Frontier Safety Framework — v2.0","authors":["Google DeepMind"],"publisher":"Google DeepMind","publicationDate":"2025-02-04","version":"Frontier Safety Framework v2.0; publication date 2025-02-04.","versionDate":"2025-02-04","canonicalUrl":"https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/updating-the-frontier-safety-framework/Frontier%20Safety%20Framework%202.0.pdf","evidenceUrl":"https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/updating-the-frontier-safety-framework/Frontier%20Safety%20Framework%202.0.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20250205011727/https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/updating-the-frontier-safety-framework/Frontier%20Safety%20Framework%202.0.pdf","archivedAt":"2025-02-05","accessedAt":"2026-09-18","reuse":{"attribution":"Google DeepMind, Frontier Safety Framework — v2.0.","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-fsf-3-1","title":"Frontier Safety Framework — v3.1","authors":["Google DeepMind"],"publisher":"Google DeepMind","publicationDate":"2026-04-17","version":"Frontier Safety Framework v3.1; publication date 2026-04-17.","versionDate":"2026-04-17","canonicalUrl":"https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/strengthening-our-frontier-safety-framework/frontier-safety-framework_3-1.pdf","evidenceUrl":"https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/strengthening-our-frontier-safety-framework/frontier-safety-framework_3-1.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20260423021420/https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/strengthening-our-frontier-safety-framework/frontier-safety-framework_3-1.pdf","archivedAt":"2026-04-23","accessedAt":"2026-09-18","reuse":{"attribution":"Google DeepMind, Frontier Safety Framework — v3.1.","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-pf-beta","title":"Preparedness Framework — Beta","authors":["OpenAI"],"publisher":"OpenAI","publicationDate":"2023-12-18","version":"Preparedness Framework Beta; publication date 2023-12-18.","versionDate":"2023-12-18","canonicalUrl":"https://cdn.openai.com/openai-preparedness-framework-beta.pdf","evidenceUrl":"https://cdn.openai.com/openai-preparedness-framework-beta.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20231218195930/https://cdn.openai.com/openai-preparedness-framework-beta.pdf","archivedAt":"2023-12-18","accessedAt":"2026-09-18","reuse":{"attribution":"OpenAI, Preparedness Framework — Beta.","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-fgf-2026","title":"Frontier Governance Framework — May 2026 edition","authors":["OpenAI"],"publisher":"OpenAI","publicationDate":"2026-05-28","version":"Frontier Governance Framework May 2026 edition; publication date 2026-05-28.","versionDate":"2026-05-28","canonicalUrl":"https://cdn.openai.com/pdf/e37d949b-8c9f-4d76-b99e-4272f4631a7e/openai-frontier-governance-framework.pdf","evidenceUrl":"https://cdn.openai.com/pdf/e37d949b-8c9f-4d76-b99e-4272f4631a7e/openai-frontier-governance-framework.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20260529171758/https://cdn.openai.com/pdf/e37d949b-8c9f-4d76-b99e-4272f4631a7e/openai-frontier-governance-framework.pdf","archivedAt":"2026-05-29","accessedAt":"2026-09-18","reuse":{"attribution":"OpenAI, Frontier Governance Framework — May 2026 edition.","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-atlas-deepmind-feedback","title":"Learning through human feedback","authors":["Jan Leike","Miljan Martic","Shane Legg"],"publisher":"DeepMind","publicationDate":"2017-06-12","version":"Dated publisher page inspected 2026-09-28; current HTML, not an authenticated historical capture.","canonicalUrl":"https://deepmind.google/blog/learning-through-human-feedback/","evidenceUrl":"https://deepmind.google/blog/learning-through-human-feedback/","language":"en","archivedUrl":"https://web.archive.org/web/20260925235348/https://deepmind.google/blog/learning-through-human-feedback/","archivedAt":"2026-09-25","accessedAt":"2026-09-28","reuse":{"attribution":"Jan Leike; Miljan Martic; Shane Legg. Learning through human feedback.","licence":"No redistribution licence asserted.","note":"Short attributed paraphrases only; original publication is linked, not redistributed."}},{"id":"source-meta-faf-v1-1","title":"Frontier AI Framework — Version 1.1","authors":["Meta"],"publisher":"Meta","publicationDate":"2025-02-03","version":"Frontier AI Framework, Version 1.1, as captured by the Internet Archive on 2025-02-13, the earliest complete capture located. The original address has returned HTTP 500 since April 2026; a revised expression under the same label was captured from 2025-03-28 and is recorded separately.","versionDate":"2025-02-03","canonicalUrl":"https://ai.meta.com/static-resource/meta-frontier-ai-framework/","evidenceUrl":"https://web.archive.org/web/20250213105620/https://ai.meta.com/static-resource/meta-frontier-ai-framework/","language":"en","archivedUrl":"https://web.archive.org/web/20250213105620/https://ai.meta.com/static-resource/meta-frontier-ai-framework/","archivedAt":"2025-02-13","accessedAt":"2026-10-03","reuse":{"attribution":"Meta, Frontier AI Framework, Version 1.1 (2025).","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-meta-aasf-v2","title":"Advanced AI Scaling Framework — Version 2","authors":["Meta"],"publisher":"Meta","publicationDate":"2026-04-08","version":"Advanced AI Scaling Framework, Version 2; its change log dates the version April 7, 2026. The file served at the recorded address is byte-identical to the Internet Archive capture of 2026-04-08.","versionDate":"2026-04-07","canonicalUrl":"https://ai.meta.com/static-resource/Meta_Advanced-AI-Scaling-Framework-v2","evidenceUrl":"https://ai.meta.com/static-resource/Meta_Advanced-AI-Scaling-Framework-v2","language":"en","archivedUrl":"https://web.archive.org/web/20260408163104/https://ai.meta.com/static-resource/Meta_Advanced-AI-Scaling-Framework-v2","archivedAt":"2026-04-08","accessedAt":"2026-10-03","reuse":{"attribution":"Meta, Advanced AI Scaling Framework, Version 2 (2026).","licence":"No open redistribution licence asserted","note":"Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice."}},{"id":"source-turing-1950-amt-b-19","title":"Computing Machinery and Intelligence (offprint, AMT/B/19)","authors":["A. M. Turing"],"publisher":"Mind","publicationDate":"1950-10","version":"Original Mind offprint, scanned by the archive; printed pp. 448–449 read on the page images for this collection (scans 16–17). Offprint held by the Turing Digital Archive, King’s College, Cambridge.","canonicalUrl":"https://turingarchive.kings.cam.ac.uk/publications-lectures-and-talks-amtb/amt-b-19","evidenceUrl":"https://turingarchive.kings.cam.ac.uk/publications-lectures-and-talks-amtb/amt-b-19","language":"en","archivedUrl":"https://web.archive.org/web/20261030045116/https://mediadepot.kings.cam.ac.uk/getmedia/81/be4a3d910d5d58e3911315113c60a09f6293fa896a5e3795998fa6b4fb72b71446336e3c7c92d5e08fd0fb2762b6f663279ba799d7fd562480b4bf63c63e02b9","archivedAt":"2026-10-30","accessedAt":"2026-10-06","reuse":{"attribution":"A. M. Turing, \"Computing Machinery and Intelligence\", Mind 59 (236), October 1950; offprint in the Turing Digital Archive, AMT/B/19.","licence":"No open redistribution licence asserted; the archive hosts the offprint with Oxford University Press’s permission. Page images are not reproduced.","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-turing-c1951-amt-b-4","title":"Intelligent Machinery, a Heretical Theory (AMT/B/4)","authors":["Alan M. Turing"],"publisher":"The Turing Digital Archive, King’s College, Cambridge","publicationDate":null,"version":"Archive scans of the carbon (p. 101, AMT/B/4/6) and the typescript (p. 10, AMT/B/4/16); the concluding paragraph read on both images for this collection. The carbon carries a typed insertion and a manuscript correction nearby.","canonicalUrl":"https://turingarchive.kings.cam.ac.uk/publications-lectures-and-talks-amtb/amt-b-4","evidenceUrl":"https://turingarchive.kings.cam.ac.uk/publications-lectures-and-talks-amtb/amt-b-4","language":"en","archivedUrl":"https://web.archive.org/web/20230511002039/https://mediadepot.kings.cam.ac.uk/getmedia/61/be4a3d910d5d58e3911315113c60a09f6293fa896a5e3795998fa6b4fb72b71446336e3c7c92d5e08fd0fb2762b6f663279ba799d7fd562480b4bf63c63e02b9","archivedAt":"2023-05-11","accessedAt":"2026-10-06","reuse":{"attribution":"Alan M. Turing, \"Intelligent Machinery, a Heretical Theory\", typescript, Turing Digital Archive AMT/B/4, King’s College, Cambridge.","licence":"No open redistribution licence asserted; archive scans are provided for research. Page images are not reproduced.","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-eoc-1997-references","title":"Engines of Creation, notes and references (web edition)","authors":["K. Eric Drexler"],"publisher":"Foresight Institute (web edition of Engines of Creation)","publicationDate":"1986","version":"Web edition page as captured by the Internet Archive on 1997-07-09 (timestamp 19970709104746); it carries the notice \"Copyright 1986, K. Eric Drexler\". Retrieved 2026-10-06.","canonicalUrl":"http://www.foresight.org/EOC/EOC_References.html","evidenceUrl":"https://web.archive.org/web/19970709104746/http://www.foresight.org:80/EOC/EOC_References.html","language":"en","archivedUrl":"https://web.archive.org/web/19970709104746/http://www.foresight.org:80/EOC/EOC_References.html","archivedAt":"1997-07-09","accessedAt":"2026-10-06","reuse":{"attribution":"K. Eric Drexler, Engines of Creation: The Coming Era of Nanotechnology (1986), Foresight Institute web edition.","licence":"All rights reserved, according to the page’s copyright notice.","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-moravec-1989-takeover-file","title":"Human Culture—A Genetic Takeover Underway","authors":["Hans Moravec"],"publisher":"Hans Moravec","publicationDate":"1989","version":"Author-hosted LaTeX file dated July 1988, retrieved 2026-10-06 (hash matched the pre-1996 candidate search); identical to the capture of 2001-03-06. Author’s project archive, Carnegie Mellon University. Date: the file’s header gives “July 1988” (the file and the capture alike); the publication date, 1989, is the year of the printed chapter in the author’s publication list. No 1989 printing was inspected.","versionDate":"1988-07","canonicalUrl":"https://frc.ri.cmu.edu/~hpm/project.archive/general.articles/1988/takeover.ltx","evidenceUrl":"https://frc.ri.cmu.edu/~hpm/project.archive/general.articles/1988/takeover.ltx","language":"en","archivedUrl":"https://web.archive.org/web/20010306111246/http://www.frc.ri.cmu.edu:80/~hpm/project.archive/general.articles/1988/takeover.ltx","archivedAt":"2001-03-06","accessedAt":"2026-10-06","reuse":{"attribution":"Hans Moravec, \"Human Culture—A Genetic Takeover Underway\", in Artificial Life (1989).","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-grace-2014-superintelligence-11","title":"Superintelligence 11: The treacherous turn","authors":["Katja Grace"],"publisher":"LessWrong","publicationDate":"2014-11-25","version":"Text checked in the post’s first stored version through the LessWrong GraphQL API on 2026-10-06; content version 1.0.0.","canonicalUrl":"https://www.lesswrong.com/posts/B39GNTsN3HocW8KFo/superintelligence-11-the-treacherous-turn","evidenceUrl":"https://www.lesswrong.com/posts/B39GNTsN3HocW8KFo/superintelligence-11-the-treacherous-turn","language":"en","archivedUrl":"https://web.archive.org/web/20180612220909/https://www.lesswrong.com/posts/B39GNTsN3HocW8KFo/superintelligence-11-the-treacherous-turn","archivedAt":"2018-06-12","accessedAt":"2026-10-06","reuse":{"attribution":"Katja Grace, \"Superintelligence 11: The treacherous turn\", LessWrong, 2014-11-25.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-lw-2026-the-talker-does-not-control-the","title":"The Talker Does Not Control The Doer (in Current AIs)","authors":["Eliezer Yudkowsky"],"publisher":"LessWrong","publicationDate":"2026-09-13","version":"Post body retrieved through the LessWrong GraphQL API on 2026-10-06 (SHA-256 172d8ecff9c433b272d41765ef05278bcbcf791a6e36550b4ad041f6f5105718), content version 1.7.0; quotations also compared with the posting-day version 1.0.0.","canonicalUrl":"https://www.lesswrong.com/posts/cJX2ssssGoYqnijwi/the-talker-does-not-control-the-doer-in-current-ais","evidenceUrl":"https://www.lesswrong.com/posts/cJX2ssssGoYqnijwi/the-talker-does-not-control-the-doer-in-current-ais","language":"en","archivedUrl":"https://web.archive.org/web/20261003174604/https://www.lesswrong.com/posts/cJX2ssssGoYqnijwi/the-talker-does-not-control-the-doer-in-current-ais","archivedAt":"2026-10-03","accessedAt":"2026-10-06","reuse":{"attribution":"Eliezer Yudkowsky, “The Talker Does Not Control The Doer (in Current AIs)”, LessWrong, 2026-09-13.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-acemoglu-restrepo-2018-ai-automation-and-work","title":"Artificial Intelligence, Automation and Work","authors":["Daron Acemoglu","Pascual Restrepo"],"publisher":"NBER Working Paper 24196","publicationDate":"2018-01","version":"Full text. Current copy retrieved directly on 2026-10-09; no archive capture (Wayback unreachable), so earlier bytes are not authenticated. SHA-256 0707870c32273416f9e92849a4eb539c74a6febbbd318e94e3d7eacab24eaea5.","canonicalUrl":"https://www.nber.org/system/files/working_papers/w24196/w24196.pdf","evidenceUrl":"https://www.nber.org/system/files/working_papers/w24196/w24196.pdf","language":"en","accessedAt":"2026-10-09","reuse":{"attribution":"Daron Acemoglu, Pascual Restrepo, “Artificial Intelligence, Automation and Work”, NBER Working Paper 24196, 2018-01.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-furman-seamans-2018-ai-and-the-economy","title":"AI and the Economy","authors":["Jason Furman","Robert Seamans"],"publisher":"NBER Working Paper 24689","publicationDate":"2018-06","version":"Full text. Current copy retrieved directly on 2026-10-09; no archive capture (Wayback unreachable), so earlier bytes are not authenticated. SHA-256 b8a5fa326ed1a114552fa25b2040e90ee7dcc15ccbbc4ee27a0fba2f43b0ef63.","canonicalUrl":"https://www.nber.org/system/files/working_papers/w24689/w24689.pdf","evidenceUrl":"https://www.nber.org/system/files/working_papers/w24689/w24689.pdf","language":"en","accessedAt":"2026-10-09","reuse":{"attribution":"Jason Furman, Robert Seamans, “AI and the Economy”, NBER Working Paper 24689, 2018-06.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-acemoglu-restrepo-2019-wrong-kind-of-ai","title":"The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand","authors":["Daron Acemoglu","Pascual Restrepo"],"publisher":"NBER Working Paper 25682","publicationDate":"2019-03","version":"Full text. Current copy retrieved directly on 2026-10-09; no archive capture (Wayback unreachable), so earlier bytes are not authenticated. SHA-256 54ad101804ab7d30ba6d136f413b1c871ffec0a273dbec290c009acd7ad32100.","canonicalUrl":"https://www.nber.org/system/files/working_papers/w25682/w25682.pdf","evidenceUrl":"https://www.nber.org/system/files/working_papers/w25682/w25682.pdf","language":"en","accessedAt":"2026-10-09","reuse":{"attribution":"Daron Acemoglu, Pascual Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand”, NBER Working Paper 25682, 2019-03.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"lg-src-cn-01ai-about","title":"关于我们 — 01.AI 零一万物","authors":["01.AI"],"publisher":"01.AI (零一万物)","publicationDate":null,"version":"Undated page, retrieved on 2026-10-15 (SHA-256 c6d11a3575e252ae47dc7548f4c1302dbd289897ab1aaf6c4d2f9b9f7829410d) directly from the publisher. Its wording can change; earlier wording is not authenticated.","canonicalUrl":"https://www.lingyiwanwu.com/about.html","evidenceUrl":"https://www.lingyiwanwu.com/about.html","language":"zh","archivedUrl":"https://web.archive.org/web/20261007110726/https://www.lingyiwanwu.com/about.html","archivedAt":"2026-10-07","accessedAt":"2026-10-15","reuse":{"attribution":"01.AI, “关于我们 — 01.AI 零一万物”.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"lg-src-cn-minimax-prospectus","title":"MiniMax Group Inc. — Global Offering","authors":["MiniMax Group Inc."],"publisher":"MiniMax","publicationDate":"2025-12-31","version":"Dated document, retrieved on 2026-10-15 (SHA-256 7853021e46da5714242f8d27984498cbfef3e260d5a7cbb8d4e6d62d83ab9df3) directly from the publisher. Publisher as printed on the inspected expression: MiniMax Group Inc. (via HKEXnews).","canonicalUrl":"https://www1.hkexnews.hk/listedco/listconews/sehk/2025/1231/2025123100025.pdf","evidenceUrl":"https://www1.hkexnews.hk/listedco/listconews/sehk/2025/1231/2025123100025.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20260930092156/https://www1.hkexnews.hk/listedco/listconews/sehk/2025/1231/2025123100025.pdf","archivedAt":"2026-09-30","accessedAt":"2026-10-15","reuse":{"attribution":"MiniMax Group Inc., “MiniMax Group Inc. — Global Offering”, 2025-12-31.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"lg-src-deal-tc-codeium-2024","title":"GitHub Copilot competitor Codeium raises $150M at a $1.25B valuation","authors":["Kyle Wiggers"],"publisher":"TechCrunch","publicationDate":"2024-08-29","version":"Dated page, retrieved on 2026-10-15 (SHA-256 71eb6c9ecc62c5f98dbbb9c355269acf29491d028241bd90f30a5d9e6e80d184) directly from the publisher.","canonicalUrl":"https://techcrunch.com/2024/08/29/github-copilot-competitor-codeium-raises-150m-at-a-1-25b-valuation/","evidenceUrl":"https://techcrunch.com/2024/08/29/github-copilot-competitor-codeium-raises-150m-at-a-1-25b-valuation/","language":"en","archivedUrl":"https://web.archive.org/web/20260923170851/https://techcrunch.com/2024/08/29/github-copilot-competitor-codeium-raises-150m-at-a-1-25b-valuation/","archivedAt":"2026-09-23","accessedAt":"2026-10-15","reuse":{"attribution":"Kyle Wiggers, “GitHub Copilot competitor Codeium raises $150M at a $1.25B valuation”, 2024-08-29.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"lg-src-deal-verge-windsurf-2025","title":"OpenAI’s Windsurf deal is off — and Windsurf’s CEO is going to Google","authors":["Hayden Field"],"publisher":"The Verge","publicationDate":"2025-07-11","version":"Dated page, retrieved on 2026-10-15 (SHA-256 adad993dc5cfd0bef51776f1a138314fc223c803d9ac0155a1e9367316399e59) directly from the publisher.","canonicalUrl":"https://www.theverge.com/openai/705999/google-windsurf-ceo-openai","evidenceUrl":"https://www.theverge.com/openai/705999/google-windsurf-ceo-openai","language":"en","archivedUrl":"https://web.archive.org/web/20261013220322/https://www.theverge.com/openai/705999/google-windsurf-ceo-openai","archivedAt":"2026-10-13","accessedAt":"2026-10-15","reuse":{"attribution":"Hayden Field, “OpenAI’s Windsurf deal is off — and Windsurf’s CEO is going to Google”, 2025-07-11.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"lg-src-gc-jiemian-wu-2025","title":"吴永辉入职字节Seed引发权力更迭？知情人士：其与朱文佳分工不同","authors":["宋佳楠"],"publisher":"界面新闻 (Jiemian)","publicationDate":"2025-02-21","version":"Dated page, retrieved on 2026-10-15 (SHA-256 f441a2bdbf927c03e2609a74b6d60b88087694ab7b2bdf4f7eb7077210ad224f) directly from the publisher.","canonicalUrl":"https://m.jiemian.com/article/12379597.html","evidenceUrl":"https://m.jiemian.com/article/12379597.html","language":"zh","archivedUrl":"https://web.archive.org/web/20250902180514/https://m.jiemian.com/article/12379597.html","archivedAt":"2025-09-02","accessedAt":"2026-10-15","reuse":{"attribution":"宋佳楠, “吴永辉入职字节Seed引发权力更迭？知情人士：其与朱文佳分工不同”, 2025-02-21.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"lg-src-deal-adept-intro-2022","title":"Introducing Adept","authors":["David Luan"],"publisher":"Adept","publicationDate":"2022-04-26","version":"Dated post, retrieved on 2026-10-15 (SHA-256 e669456e92b54c5ee26d63eda9bce65bff8dc64ced799c8b3a45fd748ae20b3c) through the third-party reader relay r.jina.ai, which renders the publisher’s current page (the publisher refused direct retrieval); not an archive copy; a fresh retrieval the same day got a block page. The current page is a later expression of the post. Archived copy: the Wayback Machine capture of 2022-04-26 16:02:16 UTC of the original address (/post/introducing-adept), titled “Introducing Adept AI Labs”, read by a cross-family machine check on 2026-10-26; no byte equivalence between the capture and the 2026 retrieval is claimed.","canonicalUrl":"https://www.adept.ai/blog/introducing-adept","evidenceUrl":"https://www.adept.ai/blog/introducing-adept","language":"en","archivedUrl":"https://web.archive.org/web/20220426160216/https://www.adept.ai/post/introducing-adept","archivedAt":"2022-04-26","accessedAt":"2026-10-15","reuse":{"attribution":"David Luan, “Introducing Adept”, 2022-04-26.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"c2-src-minimax-m3-license","title":"MiniMax-M3 LICENSE (MiniMax Community License)","authors":["MiniMax"],"publisher":"MiniMaxAI (Hugging Face)","publicationDate":"2026-06-13","version":"Dated document, retrieved on 2026-10-27 (SHA-256 b53f2fdda3049b0e9013207be51efc2d372cda1fcfdd8bb4bb8b22658ca5db9c) directly from the publisher. Version: 81b23f9add855c2eaf3d41cc38e9754897eff5f3. LICENSE at the Hugging Face commit of 13 June 2026; later edits not inspected.","canonicalUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M3/raw/81b23f9add855c2eaf3d41cc38e9754897eff5f3/LICENSE","evidenceUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M3/raw/81b23f9add855c2eaf3d41cc38e9754897eff5f3/LICENSE","language":"en","archivedUrl":"https://web.archive.org/web/20261028223920/https://huggingface.co/MiniMaxAI/MiniMax-M3/raw/81b23f9add855c2eaf3d41cc38e9754897eff5f3/LICENSE","archivedAt":"2026-10-28","accessedAt":"2026-10-27","reuse":{"attribution":"MiniMax, “MiniMax-M3 LICENSE (MiniMax Community License)”, 2026-06-13.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-cpo-orb-suncatcher-arxiv-v1","title":"Towards a future space-based, highly scalable AI infrastructure system design (arXiv:2511.19468v1)","authors":["Blaise Agüera y Arcas","Travis Beals","Maria Biggs","Jessica V. Bloom","Thomas Fischbacher","Konstantin Gromov","Urs Köster","Rishiraj Pravahan","James Manyika"],"publisher":"arXiv","publicationDate":"2025-11-22","version":"Retrieved for the candidate pack on 2026-10-28 (SHA-256 5046964f74d6036a2237fd2cd7f3ec27d5de6d336fbbbd3d5ad2a2b4d74ed1bc). arXiv version 1, stamped 22 Nov 2025. Version 2 (17 Jun 2026) was also retrieved (cache orb-suncatcher-arxiv) but is not quoted. The blog of 4 November 2025 refers to a preprint shared that day; that earlier Google-hosted copy was not retrieved. Retrieved again at admission on 2026-10-30 (SHA-256 5046964f74d6036a2237fd2cd7f3ec27d5de6d336fbbbd3d5ad2a2b4d74ed1bc, the same bytes).","canonicalUrl":"https://arxiv.org/pdf/2511.19468v1","evidenceUrl":"https://arxiv.org/pdf/2511.19468v1","language":"en","accessedAt":"2026-10-28","reuse":{"attribution":"Blaise Agüera y Arcas, Travis Beals, Maria Biggs, Jessica V. Bloom, Thomas Fischbacher, Konstantin Gromov, Urs Köster, Rishiraj Pravahan, James Manyika, “Towards a future space-based, highly scalable AI infrastructure system design (arXiv:2511.19468v1)”, 2025-11-22.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-cpo-pw-tx-sb6-2025","title":"S.B. No. 6, 89th Legislature, Regular Session (enrolled)","authors":["Texas Legislature"],"publisher":"Texas Legislature Online","publicationDate":null,"version":"Retrieved for the candidate pack on 2026-10-28 (SHA-256 1ef38096867ebc62b555037c0ab364ad002b96cd43e322f2d92686ef55f73885). Enrolled version (SB00006F) as served on 2026-10-28. Retrieved again at admission on 2026-10-30 (SHA-256 1ef38096867ebc62b555037c0ab364ad002b96cd43e322f2d92686ef55f73885, the same bytes).","canonicalUrl":"https://capitol.texas.gov/tlodocs/89R/billtext/html/SB00006F.htm","evidenceUrl":"https://capitol.texas.gov/tlodocs/89R/billtext/html/SB00006F.htm","language":"en","accessedAt":"2026-10-28","reuse":{"attribution":"Texas Legislature, “S.B. No. 6, 89th Legislature, Regular Session (enrolled)”.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-cpo-dc-nscale-uk-2025","title":"Nscale announces UK AI infrastructure commitment in partnership with Microsoft, NVIDIA and OpenAI","authors":["Nscale"],"publisher":"Nscale","publicationDate":"2025-09-16","version":"Retrieved for the candidate pack on 2026-10-28 (SHA-256 0ed575206c677a25b438171a86990da09fd960f7cfcfedcc635a7ef8792a2333). Press release page as served on 2026-10-28; dateline \"LONDON, UK, Sep 16, 2025, 10:30 PM BST\". OpenAI's \"Introducing Stargate UK\" post refuses direct retrieval. Retrieved again at admission on 2026-10-30 (SHA-256 0ed575206c677a25b438171a86990da09fd960f7cfcfedcc635a7ef8792a2333, the same bytes). Its wording can change; earlier wording is not authenticated.","canonicalUrl":"https://www.nscale.com/press-releases/nscale-uk-ai-infrastructure-announcement","evidenceUrl":"https://www.nscale.com/press-releases/nscale-uk-ai-infrastructure-announcement","language":"en","accessedAt":"2026-10-28","reuse":{"attribution":"Nscale, “Nscale announces UK AI infrastructure commitment in partnership with Microsoft, NVIDIA and OpenAI”, 2025-09-16.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}},{"id":"source-ab-gemini3-card","title":"Gemini 3 Pro Model Card","authors":["Google DeepMind"],"publisher":"Google DeepMind","publicationDate":null,"version":"Retrieved for the candidate pack on 2026-10-28 (SHA-256 9a1734bc3aa310691772d1756990b01383dc7732d4f9cb0a2888e2e06a7f84d8). PDF stating 'Model Release: November 2025, Last Updated: May 2026', as served on 2026-10-28. The deepmind.google model pages render their links in JavaScript, so the card was retrieved from the storage.googleapis.com asset host. Retrieved again at admission on 2026-10-30 (SHA-256 9a1734bc3aa310691772d1756990b01383dc7732d4f9cb0a2888e2e06a7f84d8, the same bytes).","canonicalUrl":"https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-Pro-Model-Card.pdf","evidenceUrl":"https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-Pro-Model-Card.pdf","language":"en","accessedAt":"2026-10-28","reuse":{"attribution":"Google DeepMind, “Gemini 3 Pro Model Card”.","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution."}}],"evidence":[{"id":"evidence-atlas-deepmind-feedback","sourceRole":"primary","support":"supports","locator":"Date, title and paragraph beginning “The system”","passage":"The account describes using a reward predictor to train an RL agent.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-09-28","note":"Original source passage checked by the machine agent that proposed the record. No independent machine audit or human source inspection is recorded for this increment."},{"id":"evidence-prov-rsp-2023-trusted-user-access","sourceRole":"primary","support":"supports","locator":"PDF p. 9 (printed p. 9), section 'ASL-3 Commitments', subsection 'ASL-3 Deployment Measures', bullet 'Tiered access', first sentence","passage":"In limited cases, models with capabilities relevant to catastrophic harm may be made available to a select group of vetted users with a legitimate and beneficial use-case that cannot be separated from dangerous capabilities, and only if such access can be granted safely and with sufficient oversight.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-03","note":"Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the inspected PDF, byte-identical to the Internet Archive capture of 2023-09-19 (earlier www-files.anthropic.com address). Extraction note: Quote omits the bullet's lead-in 'Tiered access:'. Human editorial review is pending."},{"id":"evidence-prov-rsp-v3-0-competitor-lead-delay","sourceRole":"primary","support":"supports","locator":"PDF p. 16 (printed p. 16), Appendix A: Commitments Related to Competitors, table row 'Anthropic in the lead', Commitment column","passage":"We will require a strong argument that catastrophic risk is contained, along the lines of our recommendations for industry-wide safety (see Section 1). We will delay AI development and deployment as needed to achieve this, until and unless we no longer believe we have a significant lead.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-03","note":"Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the PDF served through the recorded URL’s redirect, byte-identical to the Internet Archive capture of 2026-02-24. Extraction note: Scenario column of the row: Anthropic has developed or will imminently develop a highly capable model and has clear evidence that no other competitor will soon do so. Other rows: delay until meeting competitors' risk-reduction posture ('Competitors have strong safety measures'); no necessary delay ('General upleveling'). Human editorial review is pending."},{"id":"evidence-provchange-meta-aasf-v2-0-high-threshold-later","sourceRole":"primary","support":"supports","locator":"PDF p. 15, §3.3 Risk Thresholds, Table 1, High row, Risk Threshold column","passage":"Advanced AI Scaling Framework v2 states that the high risk threshold applies where deployment of the Frontier AI could substantially contribute to any threat scenario associated with a catastrophic outcome.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-03","note":"Paraphrase checked by an automated research assistant on 2026-10-03 against the cited edition; the edition’s own wording is quoted in its provision record. Human editorial review is pending."},{"id":"evidence-provchange-meta-aasf-v2-0-high-threshold-earlier","sourceRole":"primary","support":"supports","locator":"PDF p. 13, §3.3 Risk thresholds, Table 1, High row, Risk Threshold column","passage":"Frontier AI Framework v1.1 states that a model is at the high risk threshold when it provides significant uplift towards execution of a threat scenario but does not enable execution of any threat scenario identified as potentially sufficient to produce a catastrophic outcome.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-03","note":"Paraphrase checked by an automated research assistant on 2026-10-03 against the cited edition; the edition’s own wording is quoted in its provision record. Human editorial review is pending."},{"id":"evidence-ts-a3-turing-1950-context","sourceRole":"primary","support":"supports","locator":"Section 6 (5), printed p. 449 (scan 17)","passage":"It may be used to help in making up its own programmes, or to predict the effect of alterations in its own structure. By observing the results of its own behaviour it can modify its own programmes so as to achieve some purpose more effectively.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-06","note":"Read on the page image."},{"id":"evidence-ts-b6-turing-c1951","sourceRole":"primary","support":"supports","locator":"Concluding paragraph: typescript p. 10 (AMT/B/4/16, scan 16); carbon p. 101 (AMT/B/4/6, scan 6)","passage":"At some stage therefore we should have to expect the machines to take control, in the way that is mentioned in Samuel Butler's 'Erewhon'.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-06","note":"Read on the typescript page image (p. 10) for this collection and compared with the carbon (p. 101), which has the same sentence with irregular spacing. The scan has no text layer."},{"id":"evidence-ts-a1-good-1965","sourceRole":"primary","support":"supports","locator":"Section 2, \"Ultraintelligent Machines and Their Value\", printed p. 33 (scan page 3)","passage":"A machine that surpasses every intellectual activity of any person would also surpass people at designing machines; it could therefore design even better machines, and there would then be an \"intelligence explosion\", leaving human intelligence far behind.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-06","note":"Checked against the text layer of the inspected scan on 2026-10-06 (hash matched the 1996–2022 candidate search), and by a machine check on the page image of a second scan. Labelled paraphrase: the scan is a third-party copy of an in-copyright chapter."},{"id":"evidence-ts-b2-drexler-1986","sourceRole":"primary","support":"supports","locator":"Notes for chapter 5, note \"... social AI ...\" (printed p. 228 in the later electronic edition)","passage":"it would have to be able to plot and scheme, perhaps to persuade people to give it yet more information and ability.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-06","note":"Checked against the 1997 capture of the notes page on 2026-10-06; also in the later American Cryonics text."},{"id":"evidence-ts-a1-moravec-1989-context","sourceRole":"primary","support":"supports","locator":"Opening paragraphs","passage":"Machines could manage their own design and construction and continue cultural evolution without people.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-06","note":"Paraphrase checked against the retrieved file."},{"id":"evidence-ts-b5-grace-2014-treacherous-turn","sourceRole":"primary","support":"supports","locator":"Section \"Summary\", bullet citing p. 117 of Superintelligence","passage":"It is hard to observe whether an AI is dangerous via its behavior at a time when you could turn it off, because AIs have convergent instrumental reasons to pretend to be safe, even if they are not.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-06","note":"The same passage as the B2 record."},{"id":"evidence-yk-t05-e","sourceRole":"primary","support":"supports","locator":"The Talker Does Not Control The Doer, list of what 'the Grader is not', paragraph beginning 'The Grade is not pinned down'","passage":"Gradient descent on badly evaluated RL inevitably results in an AI pursuing an internalized notion of Grading where fooling the evaluator counts as winning.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-06","note":"Candidate pack passage T05-e. The quotation (24 words) matched the retrieved expression (SHA-256 172d8ecff9c433b272d41765ef05278bcbcf791a6e36550b4ad041f6f5105718) when checked on 2026-10-06. It also matched the posting-day version 1.0.0. Pack version note: edited post, current version 1.7.0; posting-day check pending."},{"id":"evidence-ec-m06a-16","sourceRole":"primary","support":"supports","locator":"pdf page 22; section 4 AI and Labor Markets","passage":"AI has the potential to continue or possibly even exacerbate trends towards declining labor force participation and increased inequality.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-14","note":"Candidate pack passage M06a-16 (research/economics-mainstream). The quotation (19 words) matched the retrieved current copy (SHA-256 b8a5fa326ed1a114552fa25b2040e90ee7dcc15ccbbc4ee27a0fba2f43b0ef63) in the pack's verification. Re-checked at admission on 2026-10-14 against a copy retrieved that day."},{"id":"lg-ev-deal-chen-cofounded-windsurf","sourceRole":"secondary","support":"supports","locator":"Paragraph 4","passage":"TechCrunch reports that Codeium was founded in 2021 by Varun Mohan together with Douglas Chen.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The paraphrase’s verbatim anchor “Codeium was founded in 2021 by Mohan” matched the retrieved text on 2026-10-15 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. The paraphrase is Selekoa’s; only the anchor is the source’s wording. A cross-family machine check (2026-10-16/17) re-read the record’s sources and agreed with the record. No human editorial review.","retrievalRoute":"direct"},{"id":"lg-ev-deal-chen-cofounder-verge-2025","sourceRole":"secondary","support":"supports","locator":"Paragraph 1","passage":"Google will instead hire Windsurf CEO Varun Mohan, cofounder Douglas Chen, and some of Windsurf’s R&D employees","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The quotation (17 words) matched the retrieved text on 2026-10-15 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-16/17) re-read the record’s sources and agreed with the record. No human editorial review.","retrievalRoute":"direct"},{"id":"lg-ev-gc-wu-yonghui-google-deepmind-2023","sourceRole":"secondary","support":"supports","locator":"Background paragraphs: the one beginning “公开资料显示” and the next, beginning “2023年”","passage":"The report states that in 2023 Google promoted Wu to Google Fellow and vice president of research at Google DeepMind.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The paraphrase’s verbatim anchor “被谷歌晋升为 “Google Fellow” 和谷歌DeepMind的研究副总裁” matched the retrieved text on 2026-10-15 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. The paraphrase is Selekoa’s; only the anchor is the source’s wording. Paraphrase of a passage not in English; no human translation review. A cross-family machine check (2026-10-16/17) re-read the record’s sources and judged it partly supported; its findings were applied at admission. No human editorial review.","retrievalRoute":"direct"},{"id":"lg-ev-cn-kai-fu-lee-01ai-founder","sourceRole":"primary","support":"supports","locator":"关于我们, 管理团队 (Management team), profile of 李开复","passage":"2023年5月，李开复博士创立 AI 2.0 大模型独角兽公司零一万物","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The quotation (4 words) matched the retrieved text on 2026-10-15 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-16/17) re-read the record’s sources and agreed with the record. No human editorial review.","retrievalRoute":"direct"},{"id":"lg-ev-deal-luan-cofounded-adept","sourceRole":"primary","support":"supports","locator":"Second body paragraph, and the paragraph beginning “True general intelligence”","passage":"Luan, the post’s author, names Ashish Vaswani and Niki Parmar as cofounders and writes that they are starting Adept.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-26","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The paraphrase’s verbatim anchor “my cofounders Ashish Vaswani and Niki Parmar” matched the text read through the third-party reader relay r.jina.ai, which renders the publisher’s current page (the publisher refused direct retrieval); not an archive copy on 2026-10-15, but a fresh retrieval the same day got a block page, so the record was held back at the genealogy admission of 2026-10-17. The passage now rests on the cross-family machine check’s read of the Wayback Machine capture of 2022-04-26 16:02:16 UTC, made on 2026-10-26 (research/archive-checks-2026-10-26/codex-report.md, item 12), which found the record supported; the capture could not be re-fetched when the record was admitted (2026-10-26). In the capture the post is titled “Introducing Adept AI Labs”, at /post/introducing-adept; the locator names the paragraph as it opens there. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. The paraphrase is Selekoa’s; only the anchor is the source’s wording. Admitted by the maintainer’s decision of 2026-10-26. No human editorial review."},{"id":"ev-cmp-rsp-v2-2-d6-competitor-1","sourceRole":"primary","support":"supports","locator":"PDF p. 17 (printed p. 13), footnote 17 (attached to §7.1 item 7 'Policy changes'), first and second sentences","passage":"It is possible at some point in the future that another actor in the frontier AI ecosystem will pass, or be on track to imminently pass, a Capability Threshold without implementing measures equivalent to the Required Safeguards","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Verbatim excerpt checked by an automated research assistant on 2026-10-27 against the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. The footnote continues that in such a scenario Anthropic might decide to lower its Required Safeguards, would acknowledge the overall risk, and would make a case to the U.S. government for regulatory action. Human editorial review is pending."},{"id":"ev-cmp-fsf-v3-1-d3-safeguards-1","sourceRole":"primary","support":"supports","locator":"PDF p. 10 (printed p. 10), §2.1.2 Deployment Mitigations, first sentence","passage":"The following mitigation process for external deployments will be applied to models reaching a misuse T/CCL","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Verbatim excerpt checked by an automated research assistant on 2026-10-27 against the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. Human editorial review is pending."},{"id":"ev-cmp-fsf-v3-1-d3-safeguards-2","sourceRole":"primary","support":"supports","locator":"PDF p. 11 (printed p. 11), §2.2 Misuse Capability Levels, first paragraph, third sentence","passage":"We recommend a security level for each CCL","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Verbatim excerpt checked by an automated research assistant on 2026-10-27 against the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. Human editorial review is pending."},{"id":"ev-cmp-preparedness-beta-2023-d3-halt-1","sourceRole":"primary","support":"supports","locator":"PDF p. 21, Governance › Safety baselines › 'Restricting deployment', first sentence","passage":"Only models with a post-mitigation score of \"medium\" or below can be deployed.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Passage also quoted in the dataset’s existing evidence for this edition, first checked on 2026-10-03. The comparison’s quotation was checked by an automated research assistant on 2026-10-27 against the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. Human editorial review is pending."},{"id":"ev-cmp-preparedness-beta-2023-d3-halt-2","sourceRole":"primary","support":"supports","locator":"PDF p. 21, Governance › Safety baselines › 'Restricting development', first sentence","passage":"Only models with a post-mitigation score of \"high\" or below can be developed further.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Passage also quoted in the dataset’s existing evidence for this edition, first checked on 2026-10-03. The comparison’s quotation was checked by an automated research assistant on 2026-10-27 against the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. Human editorial review is pending."},{"id":"evidence-ap-daron-acemoglu-acemoglu-restrepo-2018-ai-automation-and-work","sourceRole":"primary","support":"supports","locator":"PDF text layer, first two pages: author name on the title page or first page","passage":"Daron Acemoglu","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Credited name checked by an automated research assistant on 2026-10-27 in a direct retrieval and checked by a reviewer from a different model family on 2026-10-27. Human editorial review is pending.","retrievalRoute":"direct"},{"id":"evidence-ap-pascual-restrepo-acemoglu-restrepo-2019-wrong-kind-of-ai","sourceRole":"primary","support":"supports","locator":"PDF text layer, first two pages: author name on the title page or first page","passage":"Pascual Restrepo","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Credited name checked by an automated research assistant on 2026-10-27 in a direct retrieval and checked by a reviewer from a different model family on 2026-10-27. Human editorial review is pending.","retrievalRoute":"direct"},{"id":"ev-cmp-fsf-v2-0-d1-cbrn-1","sourceRole":"primary","support":"supports","locator":"PDF p. 4 (printed p. 4), Misuse Critical Capability Levels, first paragraph, first sentence","passage":"The edition says the misuse CCLs in its table were identified by continuing analysis of three domains: CBRN, Cyber and “Machine Learning R&D”.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-27","note":"Explicit paraphrase, written on 2026-10-27 by an automated research assistant after re-reading the passage in the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. The comparison quoted the passage; it is given here as a paraphrase to keep the edition within the quotation limit. Human editorial review is pending."},{"id":"ev-cmp-preparedness-beta-2023-d7-external","sourceRole":"primary","support":"supports","locator":"Preparedness · Beta: Whole text layer, PDF pp. 1–27 (all sections read).","passage":"Within the inspected scope, no passage meeting the comparison question “Public or external reporting” was found.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-26","note":"Absence recorded by an automated research assistant on 2026-10-15 for the framework comparison, audited in a fresh context, checked by a reviewer from a different model family on 2026-10-15 and re-checked on 2026-10-26. Human editorial review is pending."},{"id":"ev-cmp-frontier-governance-2026-d1-persuasion-1","sourceRole":"primary","support":"supports","locator":"PDF p. 3 (printed p. 01), §1 Introduction, second paragraph, second sentence","passage":"The edition says its processes for assessing and mitigating systemic risk span key risk categories, among them “cybersecurity”, “CBRN” (chemical, biological, radiological and nuclear), “harmful manipulation” and “loss of control risks”.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-27","note":"Explicit paraphrase, written on 2026-10-27 by an automated research assistant after re-reading the passage in the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. The comparison quoted the passage; it is given here as a paraphrase to keep the edition within the quotation limit. Human editorial review is pending."},{"id":"ev-cmp-frontier-governance-2026-d1-persuasion-2","sourceRole":"primary","support":"supports","locator":"PDF p. 11 (printed p. 09), §2.4 Risk tiers › 'Harmful manipulation', first sentence","passage":"The edition says OpenAI's approach to harmful manipulation is still exploratory.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-27","note":"Explicit paraphrase, written on 2026-10-27 by an automated research assistant after re-reading the passage in the publisher copy, whose SHA-256 matches the copy retrieved on 2026-10-15. The comparison quoted the passage; it is given here as a paraphrase to keep the edition within the quotation limit. The same sentence gives the reason as the area's newness as a systemic risk area. Human editorial review is pending."},{"id":"c2-ev-minimax-prospectus-safety-assessment","sourceRole":"primary","support":"supports","locator":"Business, AI safety and alignment measures, printed p. 279 (PDF p. 289), immediately before “Our R&D Team”","passage":"In addition, we have established a regular model safety assessment and improvement mechanism.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Candidate pack passage, Chinese frontier-AI institutions, second collection (research/china-labs-2). The quotation (13 words) matched the retrieved text on 2026-10-27 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-27, item K057) re-read the record’s sources and judged it partly supported; its corrections were applied at admission. No human editorial review.","retrievalRoute":"direct"},{"id":"c2-ev-minimax-m3-licence","sourceRole":"primary","support":"supports","locator":"LICENSE, heading","passage":"MINIMAX COMMUNITY LICENSE","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Candidate pack passage, Chinese frontier-AI institutions, second collection (research/china-labs-2). The quotation (3 words), a part of a longer passage that the pack joined across separate fields of the page, matched the retrieved text on 2026-10-27 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. The pack had joined this field with others of the same page into one quotation; at admission each field became its own quotation (cross-family check, 2026-10-27). A cross-family machine check (2026-10-27, item K060) re-read the record’s sources and agreed with the record. No human editorial review.","retrievalRoute":"direct"},{"id":"c2-ev-minimax-m3-licence-copyright","sourceRole":"primary","support":"supports","locator":"LICENSE, copyright line","passage":"Copyright (c) 2026 MiniMax","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Candidate pack passage, Chinese frontier-AI institutions, second collection (research/china-labs-2). The quotation (4 words), a part of a longer passage that the pack joined across separate fields of the page, matched the retrieved text on 2026-10-27 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. The pack had joined this field with others of the same page into one quotation; at admission each field became its own quotation (cross-family check, 2026-10-27). A cross-family machine check (2026-10-27, item K060) re-read the record’s sources and agreed with the record. No human editorial review.","retrievalRoute":"direct"},{"id":"c2-ev-minimax-m3-licence-2","sourceRole":"primary","support":"supports","locator":"LICENSE, grant","passage":"to deal in the Software for non-commercial purposes","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Candidate pack passage, Chinese frontier-AI institutions, second collection (research/china-labs-2). The quotation (8 words) matched the retrieved text on 2026-10-27 and again in a fresh retrieval the same day. Read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, apostrophes, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-27, item K060) re-read the record’s sources and agreed with the record. No human editorial review.","retrievalRoute":"direct"},{"id":"evidence-cpo-orb-oc7-suncatcher-paper-2025-1","sourceRole":"primary","support":"supports","locator":"PDF p. 5, radiation testing","passage":"In our tests, HBM-specific stress tests began to show irregularities after a cumulative dose of 2 krad(Si), which is almost 3× our minimum requirement for TID survival of 750 rad(Si).","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, compute, power and orbit (research/compute-power-orbit). The quotation (30 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-29/30, item 98) re-read the record’s sources and found full support. No human editorial review.","retrievalRoute":"direct"},{"id":"evidence-cpo-dc-stargate-uk-established-1","sourceRole":"primary","support":"supports","locator":"paragraph 4","passage":"establishing Stargate UK, an overarching infrastructure platform","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, compute, power and orbit (research/compute-power-orbit). The quotation (7 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. No human editorial review.","retrievalRoute":"direct"},{"id":"evidence-cpo-dc-stargate-uk-established-2","sourceRole":"primary","support":"supports","locator":"paragraph 4","passage":"Stargate UK is expected to be based across a number of sites in the UK, including in Cobalt Park","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, compute, power and orbit (research/compute-power-orbit). The quotation (19 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. No human editorial review.","retrievalRoute":"direct"},{"id":"evidence-cpo-dc-stargate-uk-established-3","sourceRole":"primary","support":"supports","locator":"paragraph 4","passage":"Nscale, OpenAI and NVIDIA are together establishing Stargate UK, designed to deploy OpenAI’s technology in the UK; Cobalt Park will form part of the newly designated AI Growth Zone in the North East.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, compute, power and orbit (research/compute-power-orbit). The paraphrase is Selekoa’s, written from the text retrieved for the candidate pack on 2026-10-28; it is not a quotation and was not machine-matched. The cross-family check found it faithful. No human editorial review.","retrievalRoute":"direct"},{"id":"evidence-cpo-pw-tx-sb6-75mw-threshold-1","sourceRole":"primary","support":"supports","locator":"SECTION 2, Sec. 37.0561(c)","passage":"The commission shall establish a demand threshold of 75 megawatts unless the commission determines that a lower threshold is necessary","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, compute, power and orbit (research/compute-power-orbit). The quotation (20 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-29/30, item 56) re-read the record’s sources and found full support. No human editorial review.","retrievalRoute":"direct"},{"id":"evidence-ab-ev-gemini3-card-published-1","sourceRole":"primary","support":"supports","locator":"PDF p. 2, model card preamble","passage":"Model Release: November 2025, Last Updated: May 2026","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, AI and biosecurity (research/ai-bio). The quotation (8 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-30, item 24) answered the record’s question blind from fresh retrievals and agreed in part; its corrections were applied at admission. No human editorial review.","retrievalRoute":"direct"}],"at":{"entities":[2,9,21,24,27,44,45,50,52,53,55,145,199,200,248,249,258,259,269,270,321,338,343,345,346,354,428,518,541,545,548,614,657,662,713,822,824,848,912,977,986,1077,1093,1121,1161,1181,1238],"claims":[314,402,415,513,772,774,789,811,818,940,1023,1404,1610,1629,1648,2037,2064,2114,2121,2171,2310,2600,2645,2662,2829,2836,3097,3147,3199,3401],"sources":[11,35,38,39,45,47,48,50,140,178,180,258,259,270,272,326,395,493,497,500,577,582,602,603,617,772,812,922,936,968,1085],"evidence":[336,434,447,562,563,980,982,999,1021,1029,1162,1250,1750,2021,2022,2054,2076,2635,2666,2741,2742,2750,2751,2816,2957,3265,3319,3339,3340,3610,3625,3626,3627,4073,4178,4179,4180,4339,4679]},"version":"1lg47t7"}