Review of the "Risks from automated R&D" section in the Anthropic Risk Report (February 2026)

METR reviewed Anthropic's February 2026 Risk Report section on automated R&D risks and concluded that while the report's bottom-line finding—that catastrophic risk from Claude Opus 4.6 or a less capable model automating R&D is very low—is correct, the evidence presented in the report itself is inadequate to support that conclusion. The review relied on additional non-public materials and subsequent evidence, including METR evaluations and the lack of public reports of the model automating key domains, to reach its agreement with the report's conclusion.

We reviewed the “Risks from automated R&D” section of Anthropic’s February 2026 Risk Report https://anthropic.com/feb-2026-risk-report , producing two corresponding review documents: our original review https://metr.org/assets/Original%20Review%20of%20%22Risks%20from%20automated%20R%26D%22%20section%20in%20the%20Anthropic%20Risk%20Report%20-%20February%202026.pdf and our updated review https://metr.org/assets/Updated%20Review%20of%20%22Risks%20from%20automated%20R%26D%22%20section%20in%20the%20Anthropic%20Risk%20Report%20-%20February%202026.pdf . We recommend that readers refer to our original review, which represents our review of the report as originally received. 1 https://metr.org/feed.xml fn:1 The following is the executive summary of our original review. The full documents are available as PDFs original https://metr.org/assets/Original%20Review%20of%20%22Risks%20from%20automated%20R%26D%22%20section%20in%20the%20Anthropic%20Risk%20Report%20-%20February%202026.pdf , updated https://metr.org/assets/Updated%20Review%20of%20%22Risks%20from%20automated%20R%26D%22%20section%20in%20the%20Anthropic%20Risk%20Report%20-%20February%202026.pdf . This document is METR’s external review of the “Risks from automated R&D” section in the Anthropic Risk Report: February 2026 henceforth ‘the report’ , which makes the argument that catastrophic risk from Claude Opus 4.6 or a less capable Anthropic model automating R&D in any domain is very low. Anthropic shared additional non-public materials with us for our review, and we used some non-public information shared as part of a previous review https://metr.org/blog/2026-03-12-sabotage-risk-report-opus-4-6-review/ . We further detail this process in an appendix. We lay out our findings in two sections: If we had to solely rely on the evidence presented by Anthropic in the original Risk Report, we would likely disagree with the report’s conclusion that catastrophic risk from R&D automation is very low. However, since the original release of Opus 4.6, there has been additional evidence indicating that the model is incapable of R&D in key domains, including the results of METR evaluations and the lack of public reports of the model automating any key domain. As such, we agree with the bottom-line conclusion of the report — that the risk of a catastrophe from Opus 4.6 or a less capable Anthropic model automating R&D in any domain is very low — but we think the evidence presented in the report is inadequate to establish this. We expect the public version of the report to be updated to resemble the changes anticipated in our updated review in content, though not necessarily in exact wording. We expect our updated review to cover those changes, but if the updated public version includes any changes that materially affect our opinions, we will publish a further updated review. Both documents include an appendix detailing our review process and the differences between the two versions of our review. ↩ https://metr.org/feed.xml fnref:1