Anthropic and OpenAI are negotiating with third-party AI evaluators over terms that could give both companies substantial control over which findings are made public. Anthropic CEO Dario Amodei had pledged in September 2026 to give outside evaluators ongoing, employee-like access and the freedom to publish findings, while leaving room for necessary redactions. The negotiations have reportedly shifted toward the companies seeking editorial control over what evaluators release. Anthropic named Faculty, Accenture’s AI division, as its first embedded evaluator, and key conditions under discussion include lab staff observing evaluators while they work and restrictive disclosure clauses in existing contracts that could require company consent before any publication.
Source: Read the original article
