Anthropic and OpenAI Pushed for Veto Rights Over Safety Reports While Pledging Independence

In a widely read September 2026 essay, Anthropic CEO Dario Amodei made a specific commitment that went to the heart of the embedded evaluator framework he was proposing. Outside reviewers, he wrote, should be able to publish their core findings without editorial control by Anthropic. OpenAI's Sam Altman endorsed the same principle days later, with the company publicly backing what it called "editorial independence" for outside assessors, according to TechCrunch's coverage of the pledges.The commitment mattered because the entire premise of the embedded evaluator model rests on a single assumption: that a reviewer with insider access to a lab's models, infrastructure, and internal data will produce findings the public can actually trust. Without unconditional publication rights, that assumption collapses.