Claim and evidence Close ×
Whether RSP v3.0’s new roadmap and transparency mechanisms compensate for its reduced unilateral commitments, in the policy’s overall capacity to secure risk mitigation, remains unsettled. Copy claim link
Copy claim reference
Machine-checked, not reviewed by an editor Question posed by Selekoa Unresolved
Unresolved. The recorded evidence does not settle this question. The linked evidence is background only: it neither supports nor contradicts the claim.
Context: primary + secondary · 4 sources
Basis of the change Assessed effect. An open question, and not a finding. Subject version v3.0 · Published 2026-02-24 · Effective 2026-02-24 Object version v2.2 · Publication date unknown · Effective 2025-05-14 Date 2026-02-24 · effective Scope and limits Unresolved open question, not a finding; it keeps the strength of the commitments apart from their expected effectiveness. Both sides accept the main changes: v3.0 drops the implied unilateral commitment to pause development and deployment when required safeguards cannot be met, separates recommendations for the industry from what Anthropic commits to as a company, adds transparency commitments, and introduces roadmap goals that are public goals rather than hard commitments. TIME’s report judges that overall the change leaves Anthropic far less constrained by its own safety policies. Sophie Williams and Jonas Freund, writing on GovAI’s site in a post that represents their own views, judge that it weakened future security commitments, note that existing mitigations are not lowered and conclude on balance that honesty about constraints is better than commitments that would not be followed; their article was updated on 2026-03-17 on competitor-matching commitments, and the quoted passages are unchanged. Anthropic describes the revision as amplifying what worked and committing it to more transparency. Holden Karnofsky, who led the revision and writes in his own name, expects more specific and realistic goals to improve their forcing function despite weaker consequences for missing them, and leaves the net effect uncertain. Jared Kaplan, Anthropic’s chief science officer, told TIME that unilateral commitments made no sense while competitors pressed ahead, and denied that the change was a U-turn or a capitulation to market incentives. METR’s Chris Painter, who reviewed an early draft with Anthropic’s permission, called the change understandable but was concerned it might enable a gradual “frog-boiling” effect. None of these passages settles whether the new mechanisms compensate, so all are context. The textual changes are recorded separately, and the record does not decide the question.
The evidence 14 secondary source Context only
Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ Centre for the Governance of AI (GovAI) · 2026-03-05
Archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “It has weakened future security commitments.” Location in source Section “What’s changed”, item beginning “It has weakened future security commitments”, first sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Sophie Williams; Jonas Freund Version Dated analysis by Sophie Williams and Jonas Freund; GovAI states that its research blog posts represent the views of their authors rather than the organisation. The live page records an update on 17 March 2026, after the Internet Archive capture of 2026-03-06, adding sentences on Anthropic’s commitment to match competitors’ mitigations. The four quoted passages are unchanged: they match both the live page and the capture. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ . Captured 1 day after the recorded publication date. An archive copy shows only what the archive retrieved on 2026-03-06. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Sophie Williams and Jonas Freund, GovAI, Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections (2026-03-05). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-03-06. The authors’ view (GovAI states that its research blog posts represent their authors’ views). Secondary analysis: evidence of the assessment, not of what the policy says. Recorded as context on 2026-10-04: it assesses future security commitments, not whether the new mechanisms compensate overall. Human editorial review is pending. secondary source Context only
Exclusive: Anthropic Drops Flagship Safety Pledge ↗ TIME · 2026-02-24
Archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “But overall, the change to the RSP leaves Anthropic far less constrained by its own safety policies” Location in source Paragraph beginning “But overall, the change to the RSP”, first sentence (to “safety policies”)
Machine-checked · 2026-10-03
Source details and check notes Authors Billy Perrigo Version Dated news report (published 8:00 PM UTC). Quotations match both the live page and the Internet Archive capture of 2026-02-24 20:20 UTC, the expression checked. The live page was later revised without a correction notice (structured-data modification date 2026-08-04): its third paragraph now attributes the decision to overhaul the RSP to the company over recent months rather than to CEO Dario Amodei earlier that month. The quoted passages are unchanged. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Billy Perrigo, Exclusive: Anthropic Drops Flagship Safety Pledge, TIME (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt (TIME’s own assessment, ending before the sentence’s closing clause on the previous policy) checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-02-24. Secondary reporting: evidence of the assessment, not of what the policy says. Recorded as context on 2026-10-04: it judges overall constraint, not whether the new mechanisms compensate for it. Human editorial review is pending. secondary source Context only
Exclusive: Anthropic Drops Flagship Safety Pledge ↗ TIME · 2026-02-24
Archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “Asked whether Anthropic was caving to market pressure, Kaplan argued that, in fact, Anthropic was making a renewed commitment to developing AI safely.” Location in source Paragraph beginning “Asked whether Anthropic was caving to market pressure”, first sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Billy Perrigo Version Dated news report (published 8:00 PM UTC). Quotations match both the live page and the Internet Archive capture of 2026-02-24 20:20 UTC, the expression checked. The live page was later revised without a correction notice (structured-data modification date 2026-08-04): its third paragraph now attributes the decision to overhaul the RSP to the company over recent months rather than to CEO Dario Amodei earlier that month. The quoted passages are unchanged. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Billy Perrigo, Exclusive: Anthropic Drops Flagship Safety Pledge, TIME (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt (TIME’s paraphrase of Anthropic chief science officer Jared Kaplan’s answer, not his words) checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-02-24. Reclassified as context on 2026-10-04: it answers a question about motive (market pressure), and ‘commitment’ here means dedication to developing AI safely, not a policy pledge. Human editorial review is pending. primary source Context only
Anthropic’s Responsible Scaling Policy: Version 3.0 ↗ Anthropic · 2026-02-24
Archived copy, captured 2026-02-24: Anthropic’s Responsible Scaling Policy: Version 3.0 ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “This third revision amplifies what worked about the previous RSP, commits us to more transparency about our plans and our risk considerations, and separates out our recommendations for the industry at large from what we can achieve as an individual company.” Location in source Closing section, first paragraph, second sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Anthropic Version Dated v3.0 release announcement; explicit explanation of the restructuring Canonical address Open the canonical source ↗ Accessed 2026-09-18 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Anthropic’s Responsible Scaling Policy: Version 3.0 ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-09-18.Attribution Anthropic, Responsible Scaling Policy: Version 3.0 (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-02-24. The publisher’s characterisation, not independent evidence of effect. Recorded as context on 2026-10-04: positive framing and more transparency are compatible with weaker substantive commitments. Human editorial review is pending. secondary source Context only
Exclusive: Anthropic Drops Flagship Safety Pledge ↗ TIME · 2026-02-24
Archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “I don’t think we’re making any kind of U-turn,” Kaplan says.” Location in source Paragraph beginning “Kaplan, the Anthropic executive and co-founder, denied”, final sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Billy Perrigo Version Dated news report (published 8:00 PM UTC). Quotations match both the live page and the Internet Archive capture of 2026-02-24 20:20 UTC, the expression checked. The live page was later revised without a correction notice (structured-data modification date 2026-08-04): its third paragraph now attributes the decision to overhaul the RSP to the company over recent months rather than to CEO Dario Amodei earlier that month. The quoted passages are unchanged. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Billy Perrigo, Exclusive: Anthropic Drops Flagship Safety Pledge, TIME (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt (TIME’s report of Anthropic chief science officer Jared Kaplan’s words) checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-02-24. Reclassified as context on 2026-10-04: it denies a reversal, not a weakening, and TIME frames the paragraph as Kaplan denying a capitulation to market incentives. Human editorial review is pending. secondary source Context only
Exclusive: Anthropic Drops Flagship Safety Pledge ↗ TIME · 2026-02-24
Archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “But he said he was “concerned” that moving away from binary thresholds under the previous RSP, by which the arrival of a certain capability could act as a tripwire to temporarily halt Anthropic’s AI development, might enable a “frog-boiling” effect” Location in source Paragraph beginning ““I like the emphasis on transparent risk reporting”, second sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Billy Perrigo Version Dated news report (published 8:00 PM UTC). Quotations match both the live page and the Internet Archive capture of 2026-02-24 20:20 UTC, the expression checked. The live page was later revised without a correction notice (structured-data modification date 2026-08-04): its third paragraph now attributes the decision to overhaul the RSP to the company over recent months rather than to CEO Dario Amodei earlier that month. The quoted passages are unchanged. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Billy Perrigo, Exclusive: Anthropic Drops Flagship Safety Pledge, TIME (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt (TIME’s report of METR policy director Chris Painter’s view) checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-02-24. It ends mid-sentence before TIME’s gloss of the effect. The concern is hedged (“might enable”), and TIME also reports that Painter called the change understandable and that he reviewed an early draft of the policy with Anthropic’s permission. Secondary reporting: evidence of the disagreement, not of what the policy says. Reclassified as context on 2026-10-03: a hedged concern about consequences, not a judgement that the commitments were weakened. Human editorial review is pending. secondary source Context only
Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ Centre for the Governance of AI (GovAI) · 2026-03-05
Archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “Dropping the pause commitment makes it more likely Anthropic will deploy models with unacceptable risks.” Location in source Section “Reasons to be concerned”, first sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Sophie Williams; Jonas Freund Version Dated analysis by Sophie Williams and Jonas Freund; GovAI states that its research blog posts represent the views of their authors rather than the organisation. The live page records an update on 17 March 2026, after the Internet Archive capture of 2026-03-06, adding sentences on Anthropic’s commitment to match competitors’ mitigations. The four quoted passages are unchanged: they match both the live page and the capture. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ . Captured 1 day after the recorded publication date. An archive copy shows only what the archive retrieved on 2026-03-06. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Sophie Williams and Jonas Freund, GovAI, Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections (2026-03-05). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-03-06. The authors’ view, not GovAI’s: the site states that its research blog posts represent the views of their authors rather than the organisation. Secondary analysis: evidence of the disagreement, not of what the policy says. Reclassified as context on 2026-10-03: it assesses the consequence of dropping the pause commitment rather than whether the commitments were weakened. Human editorial review is pending. secondary source Context only
Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ Centre for the Governance of AI (GovAI) · 2026-03-05
Archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “On balance, we think it’s better to be honest about constraints than to keep commitments that won’t be followed in practice.” Location in source Summary, final bullet, first sentence (repeated as the Conclusion’s final sentence)
Machine-checked · 2026-10-03
Source details and check notes Authors Sophie Williams; Jonas Freund Version Dated analysis by Sophie Williams and Jonas Freund; GovAI states that its research blog posts represent the views of their authors rather than the organisation. The live page records an update on 17 March 2026, after the Internet Archive capture of 2026-03-06, adding sentences on Anthropic’s commitment to match competitors’ mitigations. The four quoted passages are unchanged: they match both the live page and the capture. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ . Captured 1 day after the recorded publication date. An archive copy shows only what the archive retrieved on 2026-03-06. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Sophie Williams and Jonas Freund, GovAI, Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections (2026-03-05). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-03-06. The authors’ net judgement (GovAI states that its research blog posts represent their authors’ views). Context: it weighs honesty about constraints against unkept commitments and does not deny that commitments were weakened. Human editorial review is pending. secondary source Context only
Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ Centre for the Governance of AI (GovAI) · 2026-03-05
Archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “But importantly, Anthropic is not lowering any of its existing mitigations.” Location in source Summary, third bullet, final sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Sophie Williams; Jonas Freund Version Dated analysis by Sophie Williams and Jonas Freund; GovAI states that its research blog posts represent the views of their authors rather than the organisation. The live page records an update on 17 March 2026, after the Internet Archive capture of 2026-03-06, adding sentences on Anthropic’s commitment to match competitors’ mitigations. The four quoted passages are unchanged: they match both the live page and the capture. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-03-06: Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections ↗ . Captured 1 day after the recorded publication date. An archive copy shows only what the archive retrieved on 2026-03-06. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Sophie Williams and Jonas Freund, GovAI, Anthropic’s RSP v3.0: How it Works, What’s Changed, and Some Reflections (2026-03-05). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-03-06. The authors’ reading of the policy, not GovAI’s. Context: it limits the scope of any weakening to future commitments. Human editorial review is pending. primary source Context only
Anthropic’s Responsible Scaling Policy: Version 3.0 ↗ Anthropic · 2026-02-24
Archived copy, captured 2026-02-24: Anthropic’s Responsible Scaling Policy: Version 3.0 ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “The revised RSP aims to adopt more realistic unilateral commitments that are difficult but still achievable in the current environment” Location in source Section “Assessing our theory of change”, final paragraph, second sentence (to “current environment”)
Machine-checked · 2026-10-03
Source details and check notes Authors Anthropic Version Dated v3.0 release announcement; explicit explanation of the restructuring Canonical address Open the canonical source ↗ Accessed 2026-09-18 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Anthropic’s Responsible Scaling Policy: Version 3.0 ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-09-18.Attribution Anthropic, Responsible Scaling Policy: Version 3.0 (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-02-24. Anthropic’s own description of the revision: context agreed by both sides on what changed, not a judgement of whether it weakened the commitments. Human editorial review is pending. primary source Context only
Anthropic’s Responsible Scaling Policy: Version 3.0 ↗ Anthropic · 2026-02-24
Archived copy, captured 2026-02-24: Anthropic’s Responsible Scaling Policy: Version 3.0 ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “Rather than being hard commitments, these are public goals that we will openly grade our progress towards.” Location in source Section “2. Frontier Safety Roadmap”, second paragraph, first sentence
Machine-checked · 2026-10-03
Source details and check notes Authors Anthropic Version Dated v3.0 release announcement; explicit explanation of the restructuring Canonical address Open the canonical source ↗ Accessed 2026-09-18 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Anthropic’s Responsible Scaling Policy: Version 3.0 ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-09-18.Attribution Anthropic, Responsible Scaling Policy: Version 3.0 (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-03 against the live page and the Internet Archive capture of 2026-02-24. Anthropic’s own description of the revision: context agreed by both sides on what changed, not a judgement of whether it weakened the commitments. Human editorial review is pending. primary source Context only
Responsible Scaling Policy v3 ↗ LessWrong · 2026-02-24
Archived copy, captured 2026-02-24: Responsible Scaling Policy v3 ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “Even if the forcing function becomes weaker due to a less stringent “or else,” I think it will also become stronger due to the goals being more specific and realistic than the goals laid out by the old RSP.” Location in source Section “RSP v3’s attempt to amplify the good and reduce the bad”, final bullet of the list weighing the Roadmap against the old commitments, first sentence
Machine-checked · 2026-10-04
Source details and check notes Authors Holden Karnofsky Version LessWrong post by Holden Karnofsky dated 24 February 2026, 20:20 UTC. The author states that all views are his own, not Anthropic’s, and that he led the development of the new RSP. Quotations match the live page and the Internet Archive capture of 2026-02-24 20:29 UTC, taken before any comments. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Responsible Scaling Policy v3 ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Holden Karnofsky, Responsible Scaling Policy v3, LessWrong (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-04 against the live page and the Internet Archive capture of 2026-02-24 20:29 UTC. Holden Karnofsky’s own view: the post states that all views are his own, not Anthropic’s, and that he led the development of RSP v3.0. Karnofsky expects more specific and realistic goals to improve their forcing function despite weaker consequences for missing them; he leaves the net incentive effect uncertain. This does not deny weaker commitments overall. Primary evidence of his position, not independent evidence of effect. Human editorial review is pending. secondary source Context only
Exclusive: Anthropic Drops Flagship Safety Pledge ↗ TIME · 2026-02-24
Archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “We didn't really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead.” Location in source Paragraph beginning “We felt that it wouldn't actually help anyone”, second sentence
Machine-checked · 2026-10-04
Source details and check notes Authors Billy Perrigo Version Dated news report (published 8:00 PM UTC). Quotations match both the live page and the Internet Archive capture of 2026-02-24 20:20 UTC, the expression checked. The live page was later revised without a correction notice (structured-data modification date 2026-08-04): its third paragraph now attributes the decision to overhaul the RSP to the company over recent months rather than to CEO Dario Amodei earlier that month. The quoted passages are unchanged. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Exclusive: Anthropic Drops Flagship Safety Pledge ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Billy Perrigo, Exclusive: Anthropic Drops Flagship Safety Pledge, TIME (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt (TIME’s report of Anthropic chief science officer Jared Kaplan’s words) checked by an automated research assistant on 2026-10-04 against the live page and the Internet Archive capture of 2026-02-24. Context: Kaplan accepts that the revision gives up unilateral commitments, a change both sides describe; it does not decide whether the commitments taken as a whole were weakened. Human editorial review is pending. primary source Context only
Responsible Scaling Policy v3 ↗ LessWrong · 2026-02-24
Archived copy, captured 2026-02-24: Responsible Scaling Policy v3 ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “It’s removing unilateral commitments and giving more flexibility.” Location in source Q&A, “On the move away from implied unilateral commitments”, answer to the second question, first paragraph, second sentence
Machine-checked · 2026-10-04
Source details and check notes Authors Holden Karnofsky Version LessWrong post by Holden Karnofsky dated 24 February 2026, 20:20 UTC. The author states that all views are his own, not Anthropic’s, and that he led the development of the new RSP. Quotations match the live page and the Internet Archive capture of 2026-02-24 20:29 UTC, taken before any comments. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2026-02-24: Responsible Scaling Policy v3 ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2026-02-24. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution Holden Karnofsky, Responsible Scaling Policy v3, LessWrong (2026-02-24). Reuse No open redistribution licence asserted. 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. Check notes Verbatim excerpt checked by an automated research assistant on 2026-10-04 against the live page and the Internet Archive capture of 2026-02-24 20:29 UTC. Holden Karnofsky’s own view: the post states that all views are his own, not Anthropic’s, and that he led the development of RSP v3.0. Context: the author accepts that the revision removes unilateral commitments, a change both sides describe; it does not decide whether the commitments taken as a whole were weakened. Human editorial review is pending. Status and meaning Editorial status. Not reviewed by an editor . An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.
Relation. One scoped difference between document versions. The claim identifies publisher-stated change, official changelog/redline, or an explicitly inferred comparison. It does not establish implementation, effectiveness or a safety ranking.
Date. v3.0’s effective date; the assessments were published on and after that date.
Record ID rsp-v3-0-assessed-weakening