OpenAI Backs House Plan to Require Outside AI Safety Evaluators

OpenAI is backing a provision of the bipartisan FRONTIER Act in the House that would require leading AI companies to embed outside evaluators to verify their models’ safety.

The legislative support follows OpenAI’s Sept. 12 commitment to give independent evaluators access similar to that of employees, matching a pledge by Anthropic. Anthropic’s evaluators will have permanent access to its systems to verify safety practices, report incidents and assess models’ alignment during training. Both companies made those commitments as Anthropic CEO Dario Amodei called for slower advances in AI capabilities so safety work could keep up.

On Sept. 14, OpenAI CEO Sam Altman backed consistent federal safety requirements but said companies need not wait for legislation or an antitrust exemption to act. OpenAI now formulates explicit safety cases before frontier reinforcement learning runs it expects to significantly increase capabilities, in addition to checks before model releases. “When we talk about ‘pacing’, we do not mean ‘stopping’,” Altman wrote.