Reuters, citing two people familiar with the talks, reports that until at least October 5 both labs wanted strict say over what outside reviewers could publish. The concern is that findings or embedded reviewers could leak trade secrets to rivals. Ideas under discussion included having lab staff observe reviewers' teams and barring reviewers from working with two labs at once. Reuters also saw one contract that lets review firms publish confidential details without company consent. Anthropic told Reuters that reviewers it has worked with decide their own methods, findings and what to publish, but that with new firms it keeps a right to review and sign off while they learn its models. OpenAI said it does not approve whole reports, usually removes secret technical details, and lets reviewers note when large cuts were made. The reporting follows public promises last month from Dario Amodei and then OpenAI that reviewers could publish main findings unedited. For founders, third-party evaluations increasingly feed procurement and regulatory decisions, so the contract terms behind those reports affect how much weight to give them.