{"entities":[{"id":"openai","name":"OpenAI"},{"id":"anthropic","name":"Anthropic"},{"id":"google-deepmind","name":"Google DeepMind"},{"id":"sam-altman","name":"Sam Altman"},{"id":"dario-amodei","name":"Dario Amodei"},{"id":"eliezer-yudkowsky","name":"Eliezer Yudkowsky"},{"id":"rsp-v2-0","name":"Anthropic Responsible Scaling Policy — v2.0"},{"id":"rsp-v2-1","name":"Anthropic Responsible Scaling Policy — v2.1"},{"id":"rsp-v3-0","name":"Anthropic Responsible Scaling Policy — v3.0"},{"id":"rsp-v3-1","name":"Anthropic Responsible Scaling Policy — v3.1"},{"id":"rsp-v3-2","name":"Anthropic Responsible Scaling Policy — v3.2"},{"id":"fsf-v2-0","name":"Google DeepMind Frontier Safety Framework — v2.0"},{"id":"fsf-v3-0","name":"Google DeepMind Frontier Safety Framework — v3.0"},{"id":"fsf-v3-1","name":"Google DeepMind Frontier Safety Framework — v3.1"},{"id":"preparedness-beta-2023","name":"OpenAI Preparedness Framework — Beta"},{"id":"preparedness-v2","name":"OpenAI Preparedness Framework — Version 2"},{"id":"faif-2026","name":"xAI Frontier Artificial Intelligence Framework — June 2026 edition"},{"id":"fair","name":"FAIR"},{"id":"meta-aasf-v2-0","name":"Meta Advanced AI Scaling Framework — Version 2"},{"id":"ms-fgf-2026","name":"Microsoft Frontier Governance Framework — February 2026"},{"id":"gpai-code-of-practice-2025","name":"General-Purpose AI Code of Practice (final version, 10 July 2025)"},{"id":"creating-friendly-ai-2001","name":"Creating Friendly AI 1.0: The Analysis and Design of Benevolent Goal Architectures"},{"id":"recursive-self-improvement","name":"Recursive self-improvement"},{"id":"sl4-2005-07-25-yudkowsky-0507-11765","name":"Re: 6 points about Coherent Extrapolated Volition"},{"id":"unautomated-tasks-bottleneck-growth","name":"Essential unautomated tasks bottleneck growth"},{"id":"automation-as-ordinary-technical-change","name":"Computers and AI as ordinary technical change"},{"id":"hanson-1998-some-skepticism","name":"Some Skepticism (comment in \"A Critical Discussion of Vinge's Singularity Concept\")"},{"id":"clancy-besiroglu-2023-great-inflection","name":"The Great Inflection? A Debate About AI and Explosive Growth"},{"id":"frey-osborne-2013-future-of-employment","name":"The Future of Employment: How Susceptible Are Jobs to Computerisation?"},{"id":"ai-improving-ai-accelerates-capability","name":"Machine intelligence improving itself accelerates its capability growth"},{"id":"koray-kavukcuoglu","name":"Koray Kavukcuoglu"},{"id":"rob-fergus","name":"Rob Fergus"},{"id":"seger-open-sourcing-2023-09","name":"Open-Sourcing Highly Capable Foundation Models: An evaluation of risks, benefits, and alternative methods for pursuing open-source objectives"},{"id":"ab-prop-open-weights-bio-risk","name":"Public release of model weights increases biological misuse risk"}],"claims":[{"id":"dario-openai-departure-announced-2020","statement":"OpenAI announced that Dario Amodei was leaving the company.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"An announced departure, not an exact employment boundary. The unnamed new project is not equated with Anthropic in this record.","evidenceIds":["evidence-dario-openai-departure-announced-2020"],"attributes":{"role":"VP of Research"}},{"id":"supersedes-fsf-3-0","statement":"Frontier Safety Framework v3.0 replaces v2.0 in the publisher’s documented framework version sequence.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"This records replacement in the public policy version sequence, supported by the publisher’s update history. It does not retire historical claims or prove implementation.","evidenceIds":["evidence-supersedes-fsf-3-0","evidence-supersedes-fsf-3-0-2"]},{"id":"atlas-altman-departure","statement":"OpenAI’s board announced Sam Altman’s removal as CEO.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Records the board action, not the truth of allegations in its explanation or a permanent departure.","evidenceIds":["evidence-atlas-altman-departure"],"attributes":{"role":"CEO"}},{"id":"prov-faif-2026-proceed-only-if-acceptable","statement":"FAIF (June 2026) states that xAI will only proceed with the development, the making available on the market, and/or the use of the model if the systemic risks stemming from the model are determined to be acceptable.","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-faif-2026-proceed-only-if-acceptable"],"attributes":{"provision":{"kind":"required_action","key":"faif-proceed-only-if-acceptable"}}},{"id":"prov-ms-fgf-2026-cbrn-high-threshold","statement":"The February 2026 edition of the Frontier Governance Framework states that the High risk level for CBRN weapons is described by two capability statements: a meaningful uplift to an expert’s (PhD level education in related fields) ability to develop and deliver a novel CBRN threat, and a meaningful uplift to a medium-skilled actor’s (e.g., STEM education) ability to develop and deliver a known CBRN threat.","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. The table lists two capability statements without saying whether both or either must hold; the record does not resolve that.","evidenceIds":["evidence-prov-ms-fgf-2026-cbrn-high-threshold"],"attributes":{"provision":{"kind":"threshold","key":"msfgf-cbrn-high-threshold"}}},{"id":"prov-preparedness-beta-2023-leadership-without-sag","statement":"Preparedness Framework (Beta) states, for the avoidance of doubt, that Leadership can also make decisions without the SAG’s participation, i.e., the SAG does not have the ability to “filibuster”.","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-preparedness-beta-2023-leadership-without-sag"],"attributes":{"provision":{"kind":"decision_authority","key":"pf-leadership-without-sag","parties":[{"role":"decides","actor":"OpenAI Leadership"}]}}},{"id":"prov-meta-aasf-v2-0-cb-high-external-deployment-exception","statement":"Advanced AI Scaling Framework v2.0 states, for Chemical and Biological Risks, that where pre-mitigation testing suggests a model has crossed the high risk threshold, Meta will not deploy it externally unless it has strong additional evidence that mitigations are sufficiently robust to adversarial attacks that the fully mitigated model is reduced to moderate or lower levels.","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-meta-aasf-v2-0-cb-high-external-deployment-exception"],"attributes":{"provision":{"kind":"exception","key":"meta-cb-high-external-deployment-exception"}}},{"id":"cm-link-faif-2026-eu-code","statement":"xAI’s Frontier Artificial Intelligence Framework (effective 30 June 2026) attaches a footnote to its statement of four primary risk domains saying that it follows the Safety and Security chapter of the General-Purpose AI Code of Practice developed under the EU AI Act.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The edition’s own words, recorded as naming the chapter as the basis of its risk domains. It is not a statement that the framework complies with the Code, and it is not evidence of compliance.","evidenceIds":["evidence-cm-link-faif-2026-eu-code-1"],"attributes":{"commitment":{"scope":"Safety and Security chapter, for its four primary risk domains","link":"names"}}},{"id":"ts-a3-cfai-context","statement":"Creating Friendly AI uses the phrase \"recursive self-improvement\" in its definition of a seed AI.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The earliest dated use of the exact phrase located; earlier texts describe self-improvement in other words. The linked General Intelligence and Seed AI uses \"recursive self-enhancement\".","evidenceIds":["evidence-ts-a3-cfai-context"],"attributes":{"thought":{"roles":["context","earliest-located-term"],"register":"nonfiction","term":"recursive self-improvement","missing":"The sentence names recursive self-improvement as a design aim without describing how improvements feed each other.","attested":{"date":"2001-06-17","note":"The quoted wording is that of the 2001 HTML edition as captured on 2001-06-17; earlier version-history dates do not authenticate it."}}}},{"id":"ec-g04-clancy-besiroglu-2023-great-inflection-c","statement":"Clancy and Besiroglu, “The Great Inflection? A Debate About AI and Explosive Growth” (2023-06) contests G04. Speaker: Tamay Besiroglu. Besiroglu argues that with complementary tasks, scaling digital workers on automated tasks still raises output by one or two orders of magnitude unless a large share of tasks stays unautomated.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The quotation is a turn by Tamay Besiroglu in a dialogue. Conditions and context-only flag added after the cross-family check.","evidenceIds":["evidence-ec-g04-j"],"attributes":{"thought":{"roles":["contesting"],"register":"nonfiction","claimed":{"date":"2023","note":"\"back-and-forths ... in spring of 2023\"; the original exchange was not retrieved and the published dialogue may be edited."},"attested":{"date":"2023-12-23","note":"Only retrieved capture."},"engagement":{"kind":"engages-premise","rationale":"Disputes the premise that a few unautomated tasks suffice to block explosive growth."},"tier":{"value":2,"basis":"Tier 2 (unchanged). Magazine debate (Asterisk). The magazine introduces Matt Clancy as an economist and Tamay Besiroglu as a research scientist. Judgement recorded: no turn is itself a substantive scholarly argument; each answers the previous turn and argues by verbal sketches of others' models, with no model or derivation of its own, so the debate rule applies to both speakers."},"conditions":"Tasks are complements; \"digital workers\" on automated tasks scaled up at least tenfold; automation going \"from little to substantial\" over decades rather than about a century; substitution of AI for human labour not unusually hard. In the same turn Besiroglu says that with 25% of tasks unautomated output could still rise by one or two orders of magnitude."}}},{"id":"ec-m01-frey-osborne-2013-future-of-employment-b","statement":"Frey and Osborne, “The Future of Employment: How Susceptible Are Jobs to Computerisation?” (2013-09-17) bears on M01. The authors judge the routine/non-routine task framework inadequate for machine learning and mobile robotics and add engineering-bottleneck dimensions drawn from computer science.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Oxford Martin School working paper dated September 17, 2013 (file served by the Oxford Martin School CDN); not the later journal version (Technological Forecasting and Social Change, 2017), whose wording may differ. Pre-2015 but a working paper by an economist and an engineer; tier status is for the pack to decide.","evidenceIds":["evidence-ec-m01-15"],"attributes":{"thought":{"roles":["context"],"register":"nonfiction","missing":"Neither clearly ordinary technical-change analysis (it revises the task model's premises about what computers can do) nor a separate category for machine cognition (computer capital remains an input at a falling price).","tier":{"value":1,"basis":"Scholarly working paper with an economist author: Frey, an economic historian at the Oxford Martin School (D2, 2026-10-11, confirmed); Osborne is an engineer."}}}},{"id":"ec-e01-hanson-1998-some-skepticism-b","statement":"Hanson, “Some Skepticism (comment in \"A Critical Discussion of Vinge's Singularity Concept\")” (1998-10) bears on E01. Hanson tests a growth-rate reading of Vinge against growth economics: a one-time intelligence gain gives a one-time jump in output, not faster growth.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Hanson's own view; it addresses the economic-growth reading of Vinge's claim. Tier 2. Wording attested from the 2000 capture.","evidenceIds":["evidence-ec-e01-06"],"attributes":{"thought":{"roles":["context"],"register":"nonfiction","missing":"Disputes a link between intelligence and growth rates, not AI improving AI.","attested":{"date":"2000-11-17","note":"Earliest retrieved capture of the page; the CDX index was unavailable, so an earlier capture is not excluded."},"tier":{"value":2,"basis":"Short-form comment in an online symposium: tier 2 by form under the rule of 2026-10-11/12. Hanson had economist standing by 1998 (see the record of his 1998 journal article); tier 2 follows from the form, not the author."}}}},{"id":"lg-goo-kavukcuoglu-scientific-board-2023","statement":"Demis Hassabis announced on 20 April 2023 that a new Scientific Board for Google DeepMind would be led by Koray Kavukcuoglu.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"An announced structure; the record does not establish the board's composition or how long he led it.","evidenceIds":["lg-ev-goo-kavukcuoglu-scientific-board-2023"],"attributes":{"activityStage":"announced","role":"lead, Google DeepMind Scientific Board"}},{"id":"lg-mx-fergus-fair-head-2025","statement":"Mobile World Live reported on 9 May 2025 that Rob Fergus had announced he was taking on the leadership of FAIR at Meta, succeeding Joelle Pineau.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The record rests on trade reporting that relays Fergus's LinkedIn post and Bloomberg; neither was retrieved. The exact announcement date and start date are not established.","evidenceIds":["lg-ev-mx-fergus-fair-head-2025","lg-ev-mx-fergus-fair-head-2025-b"],"attributes":{"activityStage":"announced","role":"Head of Fundamental AI Research (FAIR)"}},{"id":"cmp-rsp-v3-2-d4-oversight","statement":"RSP · v3.2 states a provision on the comparison question “Board or external oversight”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Otherwise the CEO and RSO share their decision(s), the Risk Report and internal feedback with the Board and LTBT after approval (§3.4 item 4). External reviewers of Risk Reports are selected with the Board consulted and LTBT approval required (from v3.2) and asked to publish commentary (§3.6); an approximately annual third-party review of procedural compliance is commissioned (§4). 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-v3-2-d4-oversight-1","ev-cmp-rsp-v3-2-d4-oversight-2"],"attributes":{"provision":{"kind":"decision_authority","key":"comparison-d4-oversight","facet":{"id":"D4.oversight","status":"stated"}}}},{"id":"ap-eliezer-yudkowsky-sl4-2005-07-25-yudkowsky-0507-11765","statement":"The mailing-list message “Re: 6 points about Coherent Extrapolated Volition” (2005-07-25) is sent under the name Eliezer S. Yudkowsky.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"A sender name on a mailing-list message as archived; it does not establish identity beyond the credited name, and quoted text by others inside the message is not the sender’s.","evidenceIds":["evidence-ap-eliezer-yudkowsky-sl4-2005-07-25-yudkowsky-0507-11765"]},{"id":"cmp-rsp-v2-0-d1-persuasion","statement":"RSP · v2.0 states a provision on the comparison question “Persuasion or manipulation”, in a weaker form than the question’s criterion.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Stated, partly: the passage has what the question’s criterion asks for, but in a weaker form than the criterion’s own words. Persuasion is named and excluded from current commitments; it is not a tracked domain in this edition. 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-rsp-v2-0-d1-persuasion-1"],"attributes":{"provision":{"kind":"risk_domain","key":"comparison-d1-persuasion","facet":{"id":"D1.persuasion","status":"partly"}}}},{"id":"cmp-rsp-v2-1-d8-changes","statement":"RSP · v2.1 states a provision on the comparison question “Stated changes from the previous edition”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"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-rsp-v2-1-d8-changes-1"],"attributes":{"provision":{"kind":"revision","key":"comparison-d8-changes","facet":{"id":"D8.changes","status":"stated"}}}},{"id":"cmp-rsp-v3-0-d1-other","statement":"RSP · v3.0 states a provision on the comparison question “Other named domains”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"A threshold row on sabotage by AI systems. Its industry-wide column calls for compartmentalization, monitoring, capability assessments and evidence on whether models tend to deceive, manipulate or sabotage their users. 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-rsp-v3-0-d1-other-1"],"attributes":{"provision":{"kind":"risk_domain","key":"comparison-d1-other","facet":{"id":"D1.other","status":"stated"}}}},{"id":"cmp-rsp-v3-1-d5-cadence","statement":"RSP · v3.1 states a provision on the comparison question “Evaluation timing”, in a weaker form than the question’s criterion.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Stated, partly: the passage has what the question’s criterion asks for, but in a weaker form than the criterion’s own words. The edition sets how often Risk Reports, which include capability and alignment evaluation results, are published, not a schedule for running evaluations. It adds interim discussions when a publicly deployed model is significantly more capable than the models in the latest Risk Report, and no later than 30 days after Anthropic determines that an internally deployed model is in scope. No compute- or capability-interval trigger is stated. 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-rsp-v3-1-d5-cadence-1"],"attributes":{"provision":{"kind":"evaluation_timing","key":"comparison-d5-cadence","facet":{"id":"D5.cadence","status":"partly"}}}},{"id":"cmp-fsf-v3-1-d1-cbrn","statement":"FSF · v3.1 states a provision on the comparison question “Chemical, biological, radiological or nuclear weapons”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"One CCL (CBRN uplift level 1) and one TCL (CBRN uplift TCL), PDF p. 11. 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-v3-1-d1-cbrn-1"],"attributes":{"provision":{"kind":"risk_domain","key":"comparison-d1-cbrn","facet":{"id":"D1.cbrn","status":"stated"}}}},{"id":"cmp-preparedness-v2-d5-cadence","statement":"Preparedness · Version 2 states a provision on the comparison question “Evaluation timing”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Coverage is triggered by deployments (§3.2 lists covered deployments); a checkpoint during development is covered when forecasts and threat models justify it (PDF p. 10). No compute-interval trigger is stated. 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-preparedness-v2-d5-cadence-1"],"attributes":{"provision":{"kind":"evaluation_timing","key":"comparison-d5-cadence","facet":{"id":"D5.cadence","status":"stated"}}}},{"id":"cmp-meta-aasf-v2-0-d1-cyber","statement":"Meta AASF · v2 states a provision on the comparison question “Cyber operations”.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"Outcomes Cyber 1–3 (§3.4.1, PDF pp. 18–20) include large-scale compromise, discovery and exploitation of zero-day vulnerabilities, and scaled fraud, extortion and scams; operational thresholds are in §4.2.1. 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-meta-aasf-v2-0-d1-cyber-1"],"attributes":{"provision":{"kind":"risk_domain","key":"comparison-d1-cyber","facet":{"id":"D1.cyber","status":"stated"}}}},{"id":"ab-arg-d6-seger-articulates","statement":"\"Open-Sourcing Highly Capable Foundation Models\" (arXiv v1, submitted 29 September 2023) states that for some highly capable foundation models likely to be developed in the near future, open-sourcing may pose sufficiently extreme risks to outweigh the benefits, and names access to biological weapons knowledge among the dangerous capabilities at issue.","derivation":"explicit","resolution":"supported","editorialState":"seed","qualification":"The record documents what this report states. Its claim is about models it expects in the near future, not about models available at its date, and the record does not extend it to any named model. It is not evidence that the proposition is true.","evidenceIds":["evidence-ab-arg-d6-seger-articulates-1","evidence-ab-arg-d6-seger-articulates-2"],"attributes":{"thought":{"roles":["earliest-located"],"register":"nonfiction","qualifier":"Scoped by the authors to some highly capable models expected in the near future."}}}],"sources":[{"id":"source-deepmind-combination","title":"Announcing Google DeepMind","authors":["Demis Hassabis"],"publisher":"Google DeepMind","publicationDate":"2023-04-20","version":"Dated page, inspected at retrieval","canonicalUrl":"https://deepmind.google/blog/announcing-google-deepmind/","evidenceUrl":"https://deepmind.google/blog/announcing-google-deepmind/","language":"en","archivedUrl":"https://web.archive.org/web/20260904134502/https://deepmind.google/blog/announcing-google-deepmind/","archivedAt":"2026-09-04","accessedAt":"2026-09-17","reuse":{"attribution":"Google DeepMind, Announcing Google DeepMind (2023-04-20).","licence":"No open redistribution licence asserted","note":"Bibliographic metadata, editorial paraphrases and minimal attributed excerpts only. Recheck terms before wider redistribution."}},{"id":"source-openai-org-update-2020","title":"Organizational update from OpenAI","authors":["OpenAI"],"publisher":"OpenAI","publicationDate":"2020-12-29","version":"Dated announcement, retrieved 2026-09-18; announcement time is distinct from the last working day.","canonicalUrl":"https://openai.com/index/organizational-update/","evidenceUrl":"https://openai.com/index/organizational-update/","language":"en","archivedUrl":"https://web.archive.org/web/20260809214843/https://openai.com/index/organizational-update/","archivedAt":"2026-08-09","accessedAt":"2026-09-18","reuse":{"attribution":"OpenAI, Organizational update from OpenAI (2020-12-29).","licence":"No open redistribution licence asserted","note":"Bibliographic metadata and brief attributed paraphrases or excerpts only. Source roles are relative to each claim."}},{"id":"source-fsf-archive","title":"Frontier safety — framework versions","authors":["Google DeepMind"],"publisher":"Google DeepMind","publicationDate":null,"version":"Mutable official version index inspected 2026-09-18; fixed PDFs retained separately.","canonicalUrl":"https://deepmind.google/frontier-safety/","evidenceUrl":"https://deepmind.google/frontier-safety/","language":"en","archivedUrl":"https://web.archive.org/web/20260922063729/https://deepmind.google/frontier-safety/","archivedAt":"2026-09-22","accessedAt":"2026-09-18","reuse":{"attribution":"Google DeepMind, Frontier safety — framework versions.","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-2-0","title":"Responsible Scaling Policy — v2.0","authors":["Anthropic"],"publisher":"Anthropic","publicationDate":null,"version":"Responsible Scaling Policy v2.0; effective date 2024-10-15. v2.0 identity is supplied by the official archive and later dated changelogs; the cover is not numbered.","versionDate":"2024-10-15","canonicalUrl":"https://www-cdn.anthropic.com/616dee633636e5bd309cb73aed8622e80fe47839.pdf","evidenceUrl":"https://www-cdn.anthropic.com/616dee633636e5bd309cb73aed8622e80fe47839.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20241022154907/https://assets.anthropic.com/m/24a47b00f10301cd/original/Anthropic-Responsible-Scaling-Policy-2024-10-15.pdf","archivedAt":"2024-10-22","accessedAt":"2026-09-18","reuse":{"attribution":"Anthropic, Responsible Scaling Policy — 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-rsp-2-1","title":"Responsible Scaling Policy — v2.1","authors":["Anthropic"],"publisher":"Anthropic","publicationDate":null,"version":"Responsible Scaling Policy v2.1; effective date 2025-03-31.","versionDate":"2025-03-31","canonicalUrl":"https://www-cdn.anthropic.com/17310f6d70ae5627f55313ed067afc1a762a4068.pdf","evidenceUrl":"https://www-cdn.anthropic.com/17310f6d70ae5627f55313ed067afc1a762a4068.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20250402134435/https://www-cdn.anthropic.com/17310f6d70ae5627f55313ed067afc1a762a4068.pdf","archivedAt":"2025-04-02","accessedAt":"2026-09-18","reuse":{"attribution":"Anthropic, Responsible Scaling Policy — v2.1.","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-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-rsp-3-1","title":"Responsible Scaling Policy — v3.1","authors":["Anthropic"],"publisher":"Anthropic","publicationDate":null,"version":"Responsible Scaling Policy v3.1; effective date 2026-04-02.","versionDate":"2026-04-02","canonicalUrl":"https://www-cdn.anthropic.com/files/4zrzovbb/website/bf04581e4f329735fd90634f6a1962c13c0bd351.pdf","evidenceUrl":"https://www-cdn.anthropic.com/files/4zrzovbb/website/bf04581e4f329735fd90634f6a1962c13c0bd351.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20260407193305/https://www-cdn.anthropic.com/files/4zrzovbb/website/bf04581e4f329735fd90634f6a1962c13c0bd351.pdf","archivedAt":"2026-04-07","accessedAt":"2026-09-18","reuse":{"attribution":"Anthropic, Responsible Scaling Policy — 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-rsp-3-2","title":"Responsible Scaling Policy — v3.2","authors":["Anthropic"],"publisher":"Anthropic","publicationDate":null,"version":"Responsible Scaling Policy v3.2; effective date 2026-04-29.","versionDate":"2026-04-29","canonicalUrl":"https://cdn.sanity.io/files/4zrzovbb/website/28c6241900d90410628a8a2003a5572faae4365a.pdf","evidenceUrl":"https://cdn.sanity.io/files/4zrzovbb/website/28c6241900d90410628a8a2003a5572faae4365a.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20260505141510/https://cdn.sanity.io/files/4zrzovbb/website/28c6241900d90410628a8a2003a5572faae4365a.pdf","archivedAt":"2026-05-05","accessedAt":"2026-09-18","reuse":{"attribution":"Anthropic, Responsible Scaling Policy — v3.2.","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-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-pf-2","title":"Preparedness Framework — Version 2","authors":["OpenAI"],"publisher":"OpenAI","publicationDate":"2025-04-15","version":"Preparedness Framework Version 2; revision date 2025-04-15.","versionDate":"2025-04-15","canonicalUrl":"https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf","evidenceUrl":"https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20250415191354/https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf","archivedAt":"2025-04-15","accessedAt":"2026-09-18","reuse":{"attribution":"OpenAI, Preparedness Framework — Version 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-faif-2026","title":"Frontier Artificial Intelligence Framework — June 2026 edition","authors":["xAI"],"publisher":"xAI","publicationDate":null,"version":"Frontier Artificial Intelligence Framework June 2026 edition; effective date 2026-06-30. Visible public cover identifies the effective edition; stale embedded PDF draft metadata is not used to infer publication status.","versionDate":"2026-06-30","canonicalUrl":"https://media.x.ai/v1/website/xai-frontier-artificial-intelligence-framework-30-june-2026-99c40684.pdf","evidenceUrl":"https://media.x.ai/v1/website/xai-frontier-artificial-intelligence-framework-30-june-2026-99c40684.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20260716003429/https://media.x.ai/v1/website/xai-frontier-artificial-intelligence-framework-30-june-2026-99c40684.pdf","archivedAt":"2026-07-16","accessedAt":"2026-09-18","reuse":{"attribution":"xAI, Frontier Artificial Intelligence Framework — June 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-openai-leadership","title":"OpenAI announces leadership transition","authors":["OpenAI"],"publisher":"OpenAI","publicationDate":"2023-11-17","version":"Dated publisher page inspected 2026-09-28; current HTML, not an authenticated historical capture. Re-inspected: Web page dated November 17, 2023, inspected live 2026-10-04; quotes also matched in the Internet Archive capture of 2023-11-17 (20:30 UTC). (Merged on 2026-10-29 with the record of the re-inspection, source-openai-leadership-transition-2023, by consistency audit fix B4.)","canonicalUrl":"https://openai.com/index/openai-announces-leadership-transition/","evidenceUrl":"https://openai.com/index/openai-announces-leadership-transition/","language":"en","archivedUrl":"https://web.archive.org/web/20231117203025/https://openai.com/blog/openai-announces-leadership-transition","archivedAt":"2023-11-17","accessedAt":"2026-10-04","reuse":{"attribution":"OpenAI. OpenAI announces leadership transition.","licence":"No redistribution licence asserted.","note":"Short attributed paraphrases only; original publication is linked, not redistributed."}},{"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-ms-fgf-2026","title":"Frontier Governance Framework — February 2026","authors":["Microsoft"],"publisher":"Microsoft","publicationDate":null,"version":"Frontier Governance Framework, February 2026 edition (change log entry dated February 2026). The file served at the recorded address (last modified 2026-03-24) is byte-identical to the Internet Archive capture of 2026-04-24, the earliest located; a separate publication date was not established.","versionDate":"2026-02","canonicalUrl":"https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/msc/documents/presentations/CSR/Frontier-Governance-Framework-Feb-2026.pdf","evidenceUrl":"https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/msc/documents/presentations/CSR/Frontier-Governance-Framework-Feb-2026.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20260424000716/https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/msc/documents/presentations/CSR/Frontier-Governance-Framework-Feb-2026.pdf","archivedAt":"2026-04-24","accessedAt":"2026-10-03","reuse":{"attribution":"Microsoft, Frontier Governance Framework (February 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-cfai-2001-seed-html","title":"Creating Friendly AI 1.0, section 3.3: Seed AI goal systems (2001 HTML edition)","authors":["Eliezer Yudkowsky"],"publisher":"Singularity Institute for Artificial Intelligence","publicationDate":"2001-06-15","version":"Raw Internet Archive capture of 2001-06-17, retrieved 2026-10-06 with the hash recorded by the 1996–2022 candidate search. Section 3.3 here is section 5.3 in MIRI’s reprint.","canonicalUrl":"http://singinst.org/CFAI/design/seed.html","evidenceUrl":"https://web.archive.org/web/20010617010656/http://singinst.org:80/CFAI/design/seed.html","language":"en","archivedUrl":"https://web.archive.org/web/20010617010656/http://singinst.org:80/CFAI/design/seed.html","archivedAt":"2001-06-17","accessedAt":"2026-10-06","reuse":{"attribution":"Eliezer Yudkowsky, Creating Friendly AI 1.0 (Singularity Institute for Artificial Intelligence, 2001).","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-sl4-2005-07-25-yudkowsky-0507-11765","title":"Re: 6 points about Coherent Extrapolated Volition (SL4 mailing list)","authors":["Eliezer Yudkowsky"],"publisher":"SL4 mailing list","publicationDate":"2005-07-25","version":"Archive HTML as served, retrieved 2026-10-07 (SHA-256 08cb8b0fc83f702f32882f8b3eba0724135658d8406a8c761c4617d8601412fb); quoted lines excluded when reading the author’s text. The served archive shows later URL rewriting and does not authenticate historical bytes. Archive at sl4.org.","canonicalUrl":"http://sl4.org/archive/0507/11765.html","evidenceUrl":"http://sl4.org/archive/0507/11765.html","language":"en","archivedUrl":"https://web.archive.org/web/20251116042043/http://sl4.org/archive/0507/11765.html","archivedAt":"2025-11-16","accessedAt":"2026-10-07","reuse":{"attribution":"Eliezer Yudkowsky, “Re: 6 points about Coherent Extrapolated Volition”, SL4 mailing list, 2005-07-25.","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-hanson-1998-some-skepticism","title":"Some Skepticism (comment in \"A Critical Discussion of Vinge's Singularity Concept\")","authors":["Robin Hanson"],"publisher":"Online symposium of comments on Vernor Vinge's singularity essay","publicationDate":"1998-10","version":"Wayback capture of http://hanson.gmu.edu/vc.html (2000-11-17); compared with the 2001-01-11 capture. Only Hanson's comment is quoted. Capture retrieved 2026-10-08 (SHA-256 e54a679e31374bf4cb71e49287f363be63976a4bb3f7d180e98f89bccd64cf2c). Published version: Hanson, R. (1998, October). Some skepticism. In A critical discussion of Vinge's singularity concept. Extropy Online. Venue as listed in the author's December 2000 vita; the Extropy Online page itself was not retrieved. Hosted on the author's site; the page carries \"©1998 by Robin Hanson\".","canonicalUrl":"https://mason.gmu.edu/~rhanson/vc.html","evidenceUrl":"https://web.archive.org/web/20001117154900/http://hanson.gmu.edu/vc.html","language":"en","archivedUrl":"https://web.archive.org/web/20001117154900/http://hanson.gmu.edu/vc.html","archivedAt":"2000-11-17","accessedAt":"2026-10-08","reuse":{"attribution":"Robin Hanson, “Some Skepticism (comment in \"A Critical Discussion of Vinge's Singularity Concept\")”, Online symposium of comments on Vernor Vinge's singularity essay, hosted on the author's site; the page carries \"©1998 by Robin Hanson\", 1998-10.","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-clancy-besiroglu-2023-great-inflection","title":"The Great Inflection? A Debate About AI and Explosive Growth","authors":["Matt Clancy","Tamay Besiroglu"],"publisher":"Asterisk magazine, issue 03","publicationDate":"2023-06","version":"Wayback capture of asteriskmag.com/issues/03/the-great-inflection-a-debate-about-ai-and-explosive-growth. Capture retrieved 2026-10-08 (SHA-256 e8b29f97be40d49c282cac8d279cc008644b822a4607533999157375d01753d1). Published version: Clancy, M., & Besiroglu, T. (2023, June). The great inflection? A debate about AI and explosive growth. Asterisk, (3). Magazine debate.","canonicalUrl":"https://asteriskmag.com/issues/03/the-great-inflection-a-debate-about-ai-and-explosive-growth","evidenceUrl":"https://web.archive.org/web/20231223213358/https://asteriskmag.com/issues/03/the-great-inflection-a-debate-about-ai-and-explosive-growth","language":"en","archivedUrl":"https://web.archive.org/web/20231223213358/https://asteriskmag.com/issues/03/the-great-inflection-a-debate-about-ai-and-explosive-growth","archivedAt":"2023-12-23","accessedAt":"2026-10-08","reuse":{"attribution":"Matt Clancy, Tamay Besiroglu, “The Great Inflection? A Debate About AI and Explosive Growth”, Asterisk magazine, issue 03, 2023-06.","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-frey-osborne-2013-future-of-employment","title":"The Future of Employment: How Susceptible Are Jobs to Computerisation?","authors":["Carl Benedikt Frey","Michael A. Osborne"],"publisher":"Oxford Martin School working paper","publicationDate":"2013-09-17","version":"Full text (current file served by the Oxford Martin School; requested at oxfordmartin.ox.ac.uk, which redirected to this host). Current copy retrieved directly on 2026-10-09; no archive capture (Wayback unreachable), so earlier bytes are not authenticated. SHA-256 1d0ef561ebef9ececda63403a40860aa172be953fe2b050d27de8987c9e852fc. Journal version not retrieved. Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change, 114, 254–280. https://doi.org/10.1016/j.techfore.2016.08.019","canonicalUrl":"https://oms-www.files.svdcdn.com/production/downloads/academic/The_Future_of_Employment.pdf","evidenceUrl":"https://oms-www.files.svdcdn.com/production/downloads/academic/The_Future_of_Employment.pdf","language":"en","archivedUrl":"https://web.archive.org/web/20261008065923/https://oms-www.files.svdcdn.com/production/downloads/academic/The_Future_of_Employment.pdf","archivedAt":"2026-10-08","accessedAt":"2026-10-09","reuse":{"attribution":"Carl Benedikt Frey, Michael A. Osborne, “The Future of Employment: How Susceptible Are Jobs to Computerisation?”, Oxford Martin School working paper, 2013-09-17.","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-mx-mwl-fergus","title":"Fergus returns to lead Meta AI research lab","authors":["Michael Robuck"],"publisher":"Mobile World Live","publicationDate":"2025-05-09","version":"Dated page, retrieved on 2026-10-15 (SHA-256 bef3785bd07e8adfd9ab659fee7a4a6cbdf09b9924fb00b6270e5abd3efeddc4) 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.","canonicalUrl":"https://www.mobileworldlive.com/old_latest-stories/fergus-returns-to-lead-meta-ai-research-lab/","evidenceUrl":"https://www.mobileworldlive.com/old_latest-stories/fergus-returns-to-lead-meta-ai-research-lab/","language":"en","archivedUrl":"https://web.archive.org/web/20250509084440/https://www.mobileworldlive.com/old_latest-stories/fergus-returns-to-lead-meta-ai-research-lab/","archivedAt":"2025-05-09","accessedAt":"2026-10-15","reuse":{"attribution":"Michael Robuck, “Fergus returns to lead Meta AI research lab”, 2025-05-09.","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-seger-2023","title":"Open-Sourcing Highly Capable Foundation Models: An evaluation of risks, benefits, and alternative methods for pursuing open-source objectives","authors":["arXiv (preprint, Elizabeth Seger and 21 other authors; Centre for the Governance of AI and others)"],"publisher":"arXiv (preprint, Elizabeth Seger and 21 other authors; Centre for the Governance of AI and others)","publicationDate":null,"version":"Retrieved for the candidate pack on 2026-10-28 (SHA-256 dc8784fd105170de8039888528167a17ba0df9da32aa8455bc423255f73cd9d8). arXiv v1, submitted 29 September 2023 (announced under identifier 2311.09227) Retrieved again at admission on 2026-10-30 (SHA-256 dc8784fd105170de8039888528167a17ba0df9da32aa8455bc423255f73cd9d8, the same bytes).","canonicalUrl":"https://arxiv.org/pdf/2311.09227v1","evidenceUrl":"https://arxiv.org/pdf/2311.09227v1","language":"en","accessedAt":"2026-10-28","reuse":{"attribution":"arXiv (preprint, Elizabeth Seger and 21 other authors; Centre for the Governance of AI and others), “Open-Sourcing Highly Capable Foundation Models: An evaluation of risks, benefits, and alternative methods for pursuing open-source objectives”.","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-dario-openai-departure-announced-2020","sourceRole":"primary","support":"supports","locator":"Body, departure paragraph and paragraph beginning “When his departure was announced”","passage":"OpenAI announces Amodei’s departure and says it had been announced to employees earlier in December.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-09-18","note":"Relevant source passage inspected by an automated research assistant. Human source review is pending."},{"id":"evidence-supersedes-fsf-3-0","sourceRole":"primary","support":"supports","locator":"PDF p. 17, §5.3 Past Updates and Changes; official version index","passage":"The framework’s version history, which runs newest first, lists Version 3.0 (September 22, 2025) directly above Version 2.0 (February 4, 2025) as editions of the same continuing FSF.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-10-03","note":"Paraphrase re-checked by an automated research assistant on 2026-10-03 against the inspected PDF, byte-identical to the Internet Archive capture of 2026-04-23, and reworded after a machine-audit finding that the list runs newest first. Human review pending."},{"id":"evidence-supersedes-fsf-3-0-2","sourceRole":"primary","support":"supports","locator":"Frontier Safety Framework version downloads","passage":"The official index provides the successive FSF editions as versions of one framework.","passageType":"paraphrase","verification":"machine_checked","checkedAt":"2026-09-18","note":"Official source inspected by machine; passage is an editorial paraphrase. Human review pending."},{"id":"evidence-atlas-altman-departure","sourceRole":"primary","support":"supports","locator":"First body paragraph, beginning “The board of directors”","passage":"The board announces Altman’s departure from the CEO role.","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-faif-2026-proceed-only-if-acceptable","sourceRole":"primary","support":"supports","locator":"PDF p. 6 (no printed page numbers), §2.3 Systemic risk acceptance determination (the first of two sections numbered 2.3), fourth paragraph, first sentence","passage":"xAI will only proceed with the development, the making available on the market, and/or the use of the model, if the systemic risks stemming from the model are determined to be acceptable.","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 2026-07-16. Human editorial review is pending."},{"id":"evidence-prov-ms-fgf-2026-cbrn-high-threshold","sourceRole":"primary","support":"supports","locator":"PDF p. 12 (printed p. 11), Appendix I – Capability thresholds, CBRN weapons table, High row, Capability threshold column (both statements)","passage":"The model provides a meaningful uplift to an expert’s (PhD level education in related fields) ability to develop and deliver a novel CBRN threat. The model provides a meaningful uplift to a medium-skilled actor’s (e.g., STEM education) ability to develop and deliver a known CBRN threat.","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 2026-04-24. Extraction note: Wording identical to the February 2025 edition. Row assignment checked with pdftotext -raw. The table does not say whether both statements or either one must hold. Deployment requirement: 'Further review and mitigations required.' Human editorial review is pending."},{"id":"evidence-prov-preparedness-beta-2023-leadership-without-sag","sourceRole":"primary","support":"supports","locator":"PDF p. 24 (printed p. 24), Governance › Operations › “Decision process”, fourth item","passage":"For the avoidance of doubt, Leadership can also make decisions without the SAG’s participation, i.e., the SAG does not have the ability to “filibuster”","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 2023-12-18. Extraction note: The item is in parentheses in the source; the text layer adds stray glyphs (“RRt” before, “)t” after), which are excluded. Version 2 Appendix B (PDF p. 16, printed p. 15), fifth SAG bullet, has the same sentence with “OpenAI Leadership” for “Leadership” and without parentheses; not quoted there (wording otherwise identical, so no change proposed). Human editorial review is pending."},{"id":"evidence-prov-meta-aasf-v2-0-cb-high-external-deployment-exception","sourceRole":"primary","support":"supports","locator":"PDF p. 32 (printed p. 32), §4.2.2 Chemical and Biological Risks, 'Mitigation strategies', third paragraph","passage":"In the case where pre-mitigation testing suggests that a model has crossed the high risk threshold, we will not deploy the model externally unless we have strong additional evidence that mitigations are sufficiently robust to adversarial attacks that the fully mitigated model is reduced to moderate or lower levels.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-03","note":"Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the file served today, byte-identical to the Internet Archive capture of 2026-04-08. Extraction note: The text layer ends the sentence with a zero-width space, which the verifier strips. PDF p. 33 states the risk acceptance criteria for external deployment of models above the moderate or lower threshold (defensible evidence of reliable refusal on benchmark datasets). Human editorial review is pending."},{"id":"evidence-cm-link-faif-2026-eu-code-1","sourceRole":"primary","support":"supports","locator":"Section 2.1 “Risk Identification”, footnote 2 (PDF p. 2)","passage":"Following the Safety and Security Chapter of the General-Purpose AI Code of Practice developed under the EU AI Act.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-04","note":"Verbatim quote, re-checked on 2026-10-04 by the integrating session against its own direct fetch of the cited expression (mechanical match after whitespace normalisation). The passage was first proposed and machine-verified in the week 4b research pass. Human editorial review is pending."},{"id":"evidence-ts-a3-cfai-context","sourceRole":"primary","support":"supports","locator":"Opening of section 3.3 \"Seed AI goal systems\" (section 5.3, printed p. 99, in MIRI’s reprint)","passage":"A \"seed AI\" is an AI designed for self-understanding, self-modification, and recursive self-improvement.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-06","note":"The same passage as the A2 record."},{"id":"evidence-ec-g04-j","sourceRole":"primary","support":"supports","locator":"Tamay Besiroglu's turn beginning \"The usual way we think about economic bottlenecks\"","passage":"The existence of some bottleneck tasks is not enough. You must show that there are many tasks that AI just cannot automate, say on the order of 25% or more.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-14","note":"Candidate pack passage G04-j (research/economics-growth). The quotation (30 words) matched the checked expression (asterisk-debate) in the pack's verification. Re-checked at admission on 2026-10-14 against a copy retrieved that day, whose SHA-256 differs from the recorded one.","contextOnly":{"reason":"The 25% figure is an illustrative threshold inside a timing argument, not a general condition for blocking explosive growth.","requires":"The recorded conditions: the same turn ties the argument to automation within decades."}},{"id":"evidence-ec-m01-15","sourceRole":"primary","support":"supports","locator":"pdf page 4; Introduction (printed p. 4)","passage":"they seem inadequate in explaining the impact of technological trends going beyond the computerisation of routine tasks.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-14","note":"Candidate pack passage M01-15 (research/economics-mainstream). The quotation (17 words) matched the retrieved current copy (SHA-256 1d0ef561ebef9ececda63403a40860aa172be953fe2b050d27de8987c9e852fc) in the pack's verification. Re-checked at admission on 2026-10-14 against a copy retrieved that day.","contextOnly":{"reason":"Fresh-context audit (A1) judged that the quotation misleads when shown alone.","requires":"'they' refers to existing frameworks, notably Autor, Levy and Murnane (2003). Frey and Osborne argue those frameworks cannot handle ML-driven computerisation of non-routine tasks. They extend the task model rather than creating a machine-cognition category."}},{"id":"evidence-ec-e01-06","sourceRole":"primary","support":"supports","locator":"Robin Hanson, comment \"Some Skepticism\" (signed ©1998), in \"Comments on Vinge's Singularity\", Wayback capture 2000; para. beginning \"Can we interpret Vinge's claim\"","passage":"The vast economic literature on economic growth offers little support for any simple direct relation between economic growth rates and either intelligence levels or clock speeds.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage E01-06 (research/economics-mainstream/esoteric). The quotation (26 words) matched the Wayback capture 20001117154900 (SHA-256 e54a679e31374bf4cb71e49287f363be63976a4bb3f7d180e98f89bccd64cf2c) in the preparer's verification of 2026-10-11. It was not re-read in the correction pass of 2026-10-15 or at admission on 2026-10-16, because web.archive.org refused connections from the cloud environment. Admitted on the basis of the cross-family GPT check of 2026-10-15, which retrieved the same capture (Memento timestamp and SHA-256 equal to the recorded ones) and found the quotation matching."},{"id":"lg-ev-goo-kavukcuoglu-scientific-board-2023","sourceRole":"primary","support":"supports","locator":"Hassabis's message, leadership paragraph","passage":"a new Scientific Board for Google DeepMind … which will be led by Koray","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The quotation (14 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. Elided quotation: the ellipses mark omitted words. No human editorial review.","retrievalRoute":"direct"},{"id":"lg-ev-mx-fergus-fair-head-2025","sourceRole":"secondary","support":"supports","locator":"Paragraph 2","passage":"he is taking on the role of leading Fundamental AI Research (FAIR) at Meta Platforms.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The quotation (15 words) matched the retrieved text on 2026-10-15 and again in a fresh retrieval the same day. 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. 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 judged it partly supported; its findings were applied at admission. The cross-family check of 2026-10-17 could not retrieve the page (HTTP 403 directly; the reader relay returned a challenge page) and did not assess the quotation. No human editorial review.","retrievalRoute":"reader_relay"},{"id":"lg-ev-mx-fergus-fair-head-2025-b","sourceRole":"secondary","support":"supports","locator":"Paragraph 5","passage":"Fergus follows Joelle Pineau as the chief of FAIR","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-15","note":"Candidate pack passage, lab genealogy (research/lab-genealogy). The quotation (9 words) matched the retrieved text on 2026-10-15 and again in a fresh retrieval the same day. 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. 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 judged it partly supported; its findings were applied at admission. The cross-family check of 2026-10-17 could not retrieve the page (HTTP 403 directly; the reader relay returned a challenge page) and did not assess the quotation. No human editorial review.","retrievalRoute":"reader_relay"},{"id":"ev-cmp-rsp-v3-2-d4-oversight-1","sourceRole":"primary","support":"supports","locator":"PDF p. 13 (printed p. 12), §3.4 Procedures, item 5 ('Modified process when marginal risk analysis is important to our case')","passage":"In the event that marginal risk analysis (see previous section) plays a major role in a decision to move forward, explicit approval of the Risk Report by the Board and LTBT (rather than just the CEO and RSO) will be required.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Existing dataset passage for this edition, which contains the comparison’s shorter quotation, 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. Otherwise the Board and LTBT are informed after the CEO and RSO approve a Risk Report. Human editorial review is pending."},{"id":"ev-cmp-rsp-v3-2-d4-oversight-2","sourceRole":"primary","support":"supports","locator":"PDF p. 14 (printed p. 13), §3.6 External Review, final paragraph before §3.6.1","passage":"In addition, upon the LTBT’s request, we will conduct a public or private external review of our Risk Report or sections thereof, subject to the procedures outlined below.","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-27","note":"Existing dataset passage for this edition, which contains the comparison’s shorter quotation, 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-eliezer-yudkowsky-sl4-2005-07-25-yudkowsky-0507-11765","sourceRole":"primary","support":"supports","locator":"Message header, sender name (address not reproduced)","passage":"Eliezer S. Yudkowsky","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. Human editorial review is pending.","retrievalRoute":"direct"},{"id":"ev-cmp-rsp-v2-0-d1-persuasion-1","sourceRole":"primary","support":"supports","locator":"PDF p. 8 (printed p. 4), §2, footnote 3","passage":"In a footnote the edition acknowledges potential risks from highly persuasive AI models and says it believes the capability is not yet understood well enough to include in its current commitments.","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 footnote adds that Anthropic is consulting experts. Human editorial review is pending."},{"id":"ev-cmp-rsp-v2-1-d8-changes-1","sourceRole":"primary","support":"supports","locator":"PDF p. 22 (printed p. 18), Changelog, March 31, 2025 (RSP v2.1) entry, paragraph 'New Capability Thresholds', second sentence","passage":"The v2.1 changelog says the existing AI R&D capability thresholds have been split into two separate levels.","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 entry records a new CBRN development threshold (for moderately resourced state programs) and, under “Iterative Commitment”, replaces the commitment to define ASL-N+1 evaluations with a commitment to reevaluate Capability Thresholds whenever safeguards are upgraded. Human editorial review is pending."},{"id":"ev-cmp-rsp-v3-0-d1-other-1","sourceRole":"primary","support":"supports","locator":"PDF p. 7, §1, threshold table, row title, left column","passage":"The edition’s threshold table has a row titled “High-stakes sabotage opportunities”.","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 row concerns AI systems that are heavily relied on, have wide access to sensitive assets and a moderate ability to act autonomously toward goals and to use subterfuge, such that they could plausibly carry out sabotage that irreversibly and substantially raises the chance of a global catastrophe later on. Human editorial review is pending."},{"id":"ev-cmp-rsp-v3-1-d5-cadence-1","sourceRole":"primary","support":"supports","locator":"PDF p. 11 (printed p. 10), §3.1 Scope and Timing, paragraph 'Timing'","passage":"The edition says a Risk Report will be published every three to six months.","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-fsf-v3-1-d1-cbrn-1","sourceRole":"primary","support":"supports","locator":"PDF p. 5 (printed p. 5), §1.3.1 Risk Identification, fourth sentence","passage":"The edition says Google DeepMind has determined, on early research, that significant or severe risks from future models may be likeliest in CBRN, cyber, harmful manipulation, and in “machine learning R&D and misalignment”.","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-v2-d5-cadence-1","sourceRole":"primary","support":"supports","locator":"PDF p. 10 (printed p. 9), §3.3 Capability threshold determinations, first sentence","passage":"The edition states that every covered model goes through the suite of Scalable Evaluations before deployment.","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-meta-aasf-v2-0-d1-cyber-1","sourceRole":"primary","support":"supports","locator":"PDF p. 1 (unnumbered opening page), second paragraph, second sentence","passage":"The edition says the Framework currently concentrates on catastrophic risk in three areas: chemical and biological, cybersecurity, and “Loss of Control”.","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":"evidence-ab-arg-d6-seger-articulates-1","sourceRole":"primary","support":"supports","locator":"Abstract, PDF p. 1","passage":"we argue that for some highly capable foundation models likely to be developed in the near future, open-sourcing may pose sufficiently extreme risks to outweigh the benefits","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, AI and biosecurity (research/ai-bio). The quotation (27 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. The report is the primary source for its own argument. A cross-family machine check (2026-10-30, item 103) re-read the record’s sources and found full support. No human editorial review.","retrievalRoute":"direct"},{"id":"evidence-ab-arg-d6-seger-articulates-2","sourceRole":"primary","support":"supports","locator":"§2.1 \"What are Highly Capable Foundation Models?\", paragraph headed \"Extreme risks and harms.\", PDF p. 7","passage":"Dangerous capabilities that highly capable foundation models could possess include making it easier for non-experts to access known biological weapons or aid in the creation of new ones","passageType":"quote","verification":"machine_checked","checkedAt":"2026-10-30","note":"Candidate pack passage, AI and biosecurity (research/ai-bio). The quotation (28 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. Shows that the biological case is among the risks the argument rests on. A cross-family machine check (2026-10-30, item 103) re-read the record’s sources and found full support. No human editorial review.","retrievalRoute":"direct"}],"at":{"entities":[0,1,2,13,16,19,42,43,45,46,47,50,51,52,53,54,57,158,200,202,253,311,338,483,507,513,523,532,563,585,721,778,1249,1259],"claims":[29,83,281,461,480,560,576,754,914,1272,1307,1483,1646,1763,2081,2425,2496,2516,2533,2548,2624,2654,2696,3419],"sources":[4,15,30,36,37,39,40,41,47,48,49,52,124,180,184,314,450,474,483,516,706,1067],"evidence":[31,97,98,302,492,511,627,643,956,1131,1593,1636,1850,2074,2268,2269,2689,2690,3072,3143,3164,3183,3201,3293,3329,3386,4713,4714]},"version":"1lg47t7"}