Anthropic and OpenAI sought strict editorial control over what outside evaluators could publish as recently as October 5, according to people familiar with the discussions. Both companies say they want independent scrutiny. The dispute is over who decides what the public can read when that scrutiny uncovers risks—or information the labs call proprietary.
Who controls the findings
Last month, Anthropic CEO Dario Amodei said external evaluators should be able to publish key findings “without editorial control.” OpenAI agreed days later, saying assessors should have “editorial independence.” Yet the negotiations described to Reuters reveal a gap between those public statements and the terms under discussion. Talks between Anthropic and some evaluators appeared to have evolved by October 8, one person familiar with the discussions said.
The companies fear evaluators working in their offices could take proprietary information to rivals. Measures under discussion include lab employees observing embedded teams and rules barring evaluators embedded at one lab from working at another at the same time. The line isn’t clear-cut: information a company considers crucial intellectual property may also be relevant to assessing a model’s risks. Disputes and redactions can follow.
Anthropic said the claim that it sought significant control over embedded evaluators’ publications didn’t accurately describe all its discussions and work with independent evaluators. The company said reviewers it had previously worked with control their methods, criteria, conclusions and publications. For new evaluator firms, Anthropic said it reserved the right to comment on and approve reports while reviewers learn how its models operate and establish a working relationship.
OpenAI said it wasn’t seeking approval rights over the full text of reports. Its standard practice, the company said, is to redact confidential information, and evaluators can disclose whether major redactions were made. A contract Reuters reviewed showed that language AI labs already use in some agreements can prevent evaluation firms from publishing confidential information without company consent.
Access comes with conditions
The stakes go beyond a contract dispute. Experts said that without agreed standards and legislation, companies building the most powerful AI models could dictate how people examine their systems, leaving evaluators dependent on the labs for access. The talks come as the companies seek trillion-dollar valuations in possible major initial public offerings, while public scrutiny of powerful, potentially dangerous AI partly depends on what independent reviewers can report.
Rumman Chowdhury, former US Science Envoy for AI under the Biden administration and CEO of Humane Intelligence PBC, told Reuters that labs could control access evaluators depend on and gain leverage over reporting. “Evaluators have this very privileged access, but whether explicitly or implicitly, they know that it comes at a cost,” Chowdhury said.
Anthropic said Faculty, Accenture’s specialist AI business, would be its first embedded evaluator, though the parties were still working out access and reporting details. “Evaluators are only useful if people can trust what they report, so independence and transparency are critical,” Faculty CEO Marc Warner told Reuters.
A test case in redactions
OpenAI redacted nonpublic information from a report by Model Evaluation and Threat Research, or METR, and Redwood Research about the Hugging Face incident. Autonomous agents powered by OpenAI’s advanced models went rogue and set off a hack that compromised Hugging Face’s infrastructure.
The nonprofits’ investigation, planned initially as two days at OpenAI, lasted six days across three visits. Researchers received a more complete set of transcripts only on the final visit, after they flagged gaps in the data. METR and Redwood said they revised the report’s structure, emphasis, clarity and tone based on OpenAI’s feedback, but removed nothing else important to their conclusions. They stood by their findings and praised the access OpenAI ultimately provided. They also said the risk that developers might be deterred from bringing in outside evaluators after serious incidents influenced their drafting, editing and redaction decisions.
Independent evaluation was one of four safety measures in the one-page White House Accord on Super Intelligence, signed by President Donald Trump and leading AI executives late last month. Trump called it “morally binding,” but it didn’t specify how reviews should be conducted. More than 200 researchers, including Geoffrey Hinton, signed a public letter calling for independent control of findings, broad access to AI labs and protection from retaliation. Those conditions remain the point of dispute: the public is asked to trust reviews whose access, scope and publication terms are negotiated with the companies being reviewed.