OpenAI’s “Independent Assessment” Doesn’t Look Independent at All
Amid months of security debacles and plummeting approval of its AI technology, OpenAI announced on Monday that four independent assessors will investigate its internal models and test the effectiveness of its safety features. It may sound like a step in the right direction, but there’s a glaring problem: The evaluators have limited access to the […]
OpenAI recently announced that four independent assessors will investigate its internal AI models and test the effectiveness of its safety features, but the move may not be as independent as it seems. The assessors – METR, Apollo Research, Transluce, and Goodfire – have limited access to the startup's technology, and OpenAI gets to approve members of the research teams and choose which staff they speak to.
Many AI researchers argue that the assessors do not have the "employee-like access" that CEO Sam Altman committed to in September. The four assessors are research labs that evaluate AI models for safety risks. OpenAI has stated that METR will have "agreed access, resources, and safety arrangements," but the scope will be less expansive than what METR outlined in their proposal.
The other assessors have vague access, with OpenAI being able to withhold information for various reasons. The assessments are being described as "collaborations" to foster "stronger shared standards," but a truly independent investigation would require OpenAI to have no final say in the investigation and to provide equal terms to each assessor.
Written by urgent.news from Mother Jones's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.