Independent AI Safety Auditors Take Center Stage in the Industry
In the absence of federal regulation, leading AI labs like Anthropic and OpenAI are turning to independent evaluation groups to assess model capabilities and risks.
Independent safety evaluation organizations operating in the AI sector, such as METR, Apollo Research, and Transluce, are coming to the forefront as companies seek to examine model capacities and potential risks.
The Rise of Independent Groups
Operating in a relatively quiet position within the AI industry, independent evaluators are assuming a critical role to bridge security gaps at a time when the sector is growing with massive investments.
Lack of Federal Regulation
The absence of concrete federal regulatory action in the U.S. further increases the importance and weight of the mission undertaken by these third-party evaluation groups.
Corporate Approaches and Commitments
Anthropic CEO Dario Amodei announced that they would incorporate independent auditors into their ranks, a move that was also supported by OpenAI CEO Sam Altman.
Uncertainties Awaiting Resolution
Issues such as how these third-party groups will be funded, the level of access they will have to systems, and how reporting structures will be established have not yet been clarified.
Layoffs at OpenAI
OpenAI fired three employees on the grounds that they violated sensitive information access policies; two of the dismissed employees claimed that this situation was related to their communication with independent auditors.