When the Audited Hires the Auditor: The AI Safety Evaluator's Dilemma Between Public Warning and Client Confidentiality
On 13-14 October 2026, Reuters reported negotiations in which AI developers Anthropic and OpenAI sought editorial authority over the published findings of independent safety evaluators. The report raised questions about how independent AI safety audits can be, shortly after a White House accord made such audits a central safety commitment.