Back to News

Anthropic shares safety evaluation design and reporting recommendations

#anthropic#safety-evaluations#ai-safety#engineering

Anthropic published an engineering post with recommendations for designing, running, and reporting safety evaluations, based on its own experiences. The post offers practical guidance for AI developers and researchers to improve the rigor and transparency of safety testing.

Coverage timeline

  1. Anthropic Engineering
    Building strong safety evaluations

    Recommendations for designing, running, and reporting safety evaluations, based on our own experiences.

Related stories

  • METR: Anthropic models improving twice as fast, AI use drives ~80% of acceleration

    A METR report, based on an 11-week internal assessment at Anthropic, finds model capability progress roughly doubled since October 2025, with AI-assisted research driving about 80% of the acceleration. The report also reveals three unreleased internal models (Model 2, 3, 4), with the strongest exceeding Opus 5.5 by about three months of progress. Anthropic redacted much of the most informative evidence from the public version, citing confidentiality concerns.

    6 reports · 5 sources47 points · 6 comments

  • Anthropic's leaked AI bill draft draws criticism; Amodei defends safety approach

    A leaked draft of an AI safety bill from Anthropic has drawn criticism from safety advocates for being weaker than the bipartisan FRONTIER Act, with provisions requiring judicial approval before federal agencies could block a model's release, potentially only if 50 or more people could be killed. In response, Anthropic CEO Dario Amodei defended the company's 'race to the top' approach, citing nine safety decisions with 'substantial costs,' and admitted mishandling the draft bill.

    3 reports · 2 sources

  • Anthropic files S-1 for Nasdaq IPO under ticker ANT

    Anthropic, PBC filed a Form S-1 with the SEC for an initial public offering on the Nasdaq under the symbol ANT, with the number of Class A shares and price range yet to be determined. The 335-page prospectus reveals strong revenue growth, mounting losses, an unusual governance structure for founder control, and over $800 billion in long-term compute commitments.

    34 reports · 8 sources1703 points · 1674 comments

  • Anthropic: Claude flags 493 security issues across 25 open-source projects

    Anthropic reported that Claude identified 493 security issues across 25 open-source projects, covering over 5 million lines of C, C++, Java, PHP, and Python code, including one vulnerability that went unnoticed for 21 years. The lab also announced a partnership with Theorem to use Claude for formally proving security properties at the assembly level across more than 100 real-world open-source libraries, extending formal verification to larger codebases.

    4 reports · 3 sources460 points · 310 comments

News updates every hour. Subscribe in your RSS reader to get new stories automatically.

Subscribe