Reuters: OpenAI and Anthropic sought control over what outside AI evaluators can publish

Reuters reported on 13 October 2026 that Anthropic and OpenAI have sought significant control over what outside AI safety evaluators can publish, fearing trade-secret leaks — even as both labs publicly pledged evaluator independence last month.

What Reuters reported

Two people familiar with the discussions said that as recently as 5 October the companies were seeking strict editorial control over evaluator publications; by 8 October, talks with some Anthropic evaluators appeared to have evolved. Reuters also described measures under discussion such as lab staff observing embedded evaluator teams and barring simultaneous work at rival labs.

Company responses

Anthropic told Reuters that seeking significant control “did not accurately characterize” all of its work with independent evaluators. With returning partners, it said reviewers control methods, criteria, conclusions and what to publish; for new evaluator firms it reserves the right to comment on and approve reports while reviewers learn its models.

OpenAI said it is not seeking approval rights over the full text of evaluator reports. Its standard practice, it said, is to redact confidential information, and evaluators can disclose whether major redactions were made.

Why it matters

Independent evaluation is one pillar of the White House Accord on Super Intelligence, but the accord does not specify how reviews work. Faculty CEO Marc Warner told Reuters independence and transparency are critical for trust. Rumman Chowdhury warned that without standards, labs that control access also hold leverage over what gets reported. Reuters also noted OpenAI redacted nonpublic material from a METR/Redwood report on a Hugging Face agent incident; the nonprofits said they edited structure and tone after feedback but stood by their conclusions.

The Verge summarized the same Reuters exclusive.