# Anthropic and OpenAI Sought Control Over What Outside AI Evaluators Publish, Reuters Reports; Both Labs Dispute Parts of the Account

Times of AI Desk · 2026-10-13 · Policy

[https://timesof.ai/2026/10/anthropic-openai-evaluator-publication-control-reuters](https://timesof.ai/2026/10/anthropic-openai-evaluator-publication-control-reuters)

> Reuters, citing two people familiar with the talks, reports that Anthropic and OpenAI sought significant control over what outside evaluators can publish, out of concern that findings could reveal trade secrets. Anthropic says that does not describe all its work and that returning evaluators control what they publish; OpenAI says it is not seeking approval rights over full reports.

**Anthropic** and **OpenAI** have sought significant control over what outside evaluators can publish, out of concern that findings could reveal trade secrets, **Reuters** reported on **Tuesday, October 13**, citing two people familiar with the discussions. Both companies dispute parts of that account. Last month Anthropic CEO **Dario Amodei** said evaluators should be able to publish key findings "without editorial control," and OpenAI said days later that assessors should have "editorial independence," Reuters reports.

## What each lab says

Anthropic told Reuters that seeking significant control over embedded evaluators' publications "did not accurately characterize" all of its discussions and work with independent evaluators, as Reuters paraphrases it. With evaluators it has worked with before, Anthropic said it had made clear that they control their methods, evaluation criteria, conclusions and what to publish. For new evaluator firms, Anthropic said it reserves the right to comment on and approve reports while the reviewers learn how its models work.

OpenAI told Reuters it is not seeking approval rights over the full text of evaluators' reports. It said its standard practice is to redact confidential information, and that evaluators can disclose whether major redactions were made.

## What Reuters' sources describe

As recently as **October 5**, both labs were seeking strict editorial control over what evaluators could publish, the two people told Reuters. By **October 8**, talks between Anthropic and some evaluators "appeared to have evolved," according to one of them.

The labs are worried that evaluator staff working inside their walls could take proprietary information to rivals, the people said. Measures under discussion include having lab employees observe embedded teams and barring evaluators embedded at one lab from working at another at the same time, one person told Reuters. A separate person with knowledge of past dealings said that what labs treat as intellectual property can overlap with information relevant to judging a model's risk, which leads to arguments and redactions.

Reuters also says it saw a contract whose language can bar an evaluation firm from publishing confidential information without the lab's consent.

## Evaluators and outside voices

Anthropic named Accenture's **Faculty** as its first embedded evaluator last month, with access and reporting rules still being worked out, Reuters notes; we covered that [Anthropic–Accenture arrangement](https://timesof.ai/2026/09/anthropic-accenture-embedded-evaluation) at the time. "Evaluators are only useful if people can trust what they report, so independence and transparency are critical," Faculty CEO **Marc Warner** told Reuters.

**Rumman Chowdhury**, CEO of Humane Intelligence, told Reuters that without standards the labs control the access evaluators depend on. "Evaluators have this very privileged access, but whether explicitly or implicitly, they know that it comes at a cost," she said.

Reuters points to two recent texts. Independent evaluation is one of four safety measures in the White House accord signed late last month, [which we covered](https://timesof.ai/2026/09/white-house-frontier-responsibilities-accord) under its published title, the Joint Commitment on Frontier Responsibilities; Reuters notes the accord does not say how reviews should be run. A public letter from evaluators, signed by more than 200 researchers including **Geoffrey Hinton**, calls for independent control of findings, broad access and protection from retaliation.

## The Hugging Face report as a precedent

OpenAI redacted nonpublic information from the METR and Redwood Research report on the Hugging Face incident, Reuters reports; we covered [that report](https://timesof.ai/2026/08/openai-hugging-face-incident-technical-report) in August. The investigation, first planned as two days at OpenAI, took six days over three visits, according to Reuters. METR and Redwood said they edited the report's structure, emphasis, clarity and tone after OpenAI's feedback, that nothing else important to their conclusions was removed, and that they stand by their findings, Reuters reports.

## Limits

- The core account comes from two unnamed people and one Reuters-seen contract; we could not check it independently. The Next Web's write-up relies on Reuters.
- Anthropic's and OpenAI's positions are as they gave them to Reuters. The talks were still changing as of Reuters' reporting.

## Sources

- [Reuters via Investing.com: Exclusive: Behind AI industry's call for independent reviews, a fight over what gets made public (October 13, 2026)](https://www.investing.com/news/stock-market-news/exclusivebehind-ai-industrys-call-for-independent-reviews-a-fight-over-what-gets-made-public-4945991)
- [The Next Web: Anthropic and OpenAI evaluators publication control (October 13, 2026)](https://thenextweb.com/news/anthropic-openai-ai-evaluators-publication-control)
- [Times of AI: Anthropic and Accenture embedded evaluation](https://timesof.ai/2026/09/anthropic-accenture-embedded-evaluation)
- [Washington Examiner: full text of the White House Accord on Super Intelligence / Joint Commitment (September 29, 2026)](https://www.washingtonexaminer.com/news/white-house/4747747/full-trump-white-house-accord-ai-super-intelligence/)
- [Times of AI: Trump and Frontier CEOs Sign a Voluntary ‘Morally Binding’ Accord](https://timesof.ai/2026/09/white-house-frontier-responsibilities-accord)
- [Times of AI: Hugging Face incident technical report](https://timesof.ai/2026/08/openai-hugging-face-incident-technical-report)