Anthropic grants safety evaluators inside access to frontier AI models
Anthropic is providing non-profit evaluation groups, including METR and SecureBio, with employee-like access to its models to assess artificial intelligence risks and safety practices.
Anthropic has begun granting non-profit evaluation group METR employee-like access to its frontier artificial intelligence models, fulfilling a commitment made last month to expand independent, third-party assessments of its systems and safety practices.[1][3][4][6]
The initiative also includes non-profit SecureBio, according to CNBC, as Anthropic enlists external evaluators to help monitor safety practices and model risks. Crypto Briefing noted that while the collaboration aims to enhance transparency, redaction policies could influence public trust in the evaluations.[2][5]
Key facts
- Anthropic stated it is granting employee-like access to METR as part of a pledge to partner with third-party evaluators to assess its models and safety practices.
- A team from METR was granted employee-like access specifically to Anthropic's frontier models.
- Anthropic also engaged non-profit SecureBio alongside METR to assess risks and safety protocols.
- Anthropic committed the prior month to strengthening independent, third-party assessments.
Sources · 6 sources
- MT
MTS@MTSlivePost on X ·
SITUATION DETECTED: Anthropic has started giving employee-like access to their frontier models to a team from METR.
Open source - CN
CNBC@CNBCPost on X ·
Anthropic taps nonprofits METR and SecureBio to help assess risks and safety practices https://t.co/iiehqbsBXu
Open source - TE
TechmemeArticle ·
Anthropic says it is giving employee-like access to METR as part of its pledge to partner with evaluators for assessment of its models and safety practices (Anthropic) Anthropic : Anthropic says it is giving employee-like access to METR as part of its pledge to partner with evaluators for assessment of its models and safety practices — Last month, we committed to stronger independent, third-party assessment of our models and safety practices.
Open source - TE
Techmeme@TechmemePost on X ·
Anthropic says it is giving employee-like access to METR as part of its pledge to partner with evaluators for assessment of its models and safety practices (Anthropic) (Visit Techmeme dot com for the link and full context!)
Open source - CB
Crypto BriefingArticle ·
Anthropic gives outside AI safety evaluators METR and SecureBio inside access Anthropic's collaboration with external evaluators could enhance AI safety transparency, but redaction policies may impact public trust. The post Anthropic gives outside AI safety evaluators METR and SecureBio inside access appeared first on Crypto Briefing .
Open source - FT
Financial Times@FTPost on X ·
Anthropic grants non-profit METR ‘employee-like’ access to monitor AI risks https://t.co/n5LZs1x9Em
Open source