AI Experts Demand Independent Safety Audits for Anthropic, OpenAI

More than 100 artificial intelligence experts have signed a public letter demanding that Anthropic, OpenAI, and other major foundation model laboratories submit to truly independent safety evaluations, warning that self-governed assessments are insufficient as AI systems grow more powerful.

The coalition's core argument centers on a structural conflict of interest: when AI labs conduct or commission their own safety evaluations, there is little guarantee that findings will be reported fully or acted upon. Experts in the letter are pushing for evaluators who operate without financial or institutional ties to the companies they are assessing.

Read more Trump Plans Major US Military Presence in Greenland Security Deal →

Transparency is the second major pillar of the experts' demands. Signatories argue that evaluation methodologies, results, and red-team findings should be made available to the broader research community and policymakers — not kept internal where they could be selectively disclosed or quietly shelved.

The letter arrives at a pivotal moment for AI governance. Anthropic and OpenAI are among the most closely watched companies in technology, with both having made public commitments to safety while simultaneously racing to deploy increasingly capable frontier models. Critics have long questioned whether voluntary safety pledges carry enough weight without external accountability structures to enforce them.

The push for independent oversight mirrors debates already unfolding in Washington and Brussels, where regulators are exploring mandatory audit requirements for high-risk AI systems. Whether the letter translates into concrete policy pressure on the labs remains to be seen, but the broad expert consensus it represents adds meaningful weight to calls for structural reform in how AI safety is verified. Continue reading at US Top News and Analysis.