What happened
According to Reuters, Anthropic and OpenAI sought a significant say over what outside evaluators of their AI models can make public. The report cites two people familiar with the talks.
The companies' rationale
Both firms worry that findings from outside reviews could expose their trade secrets. That creates tension between the transparency of independent evaluation and the companies' commercial interests.
Promise versus practice
According to the report, both labs had recently promised outside reviewers a free hand. This new information raises questions about how those commitments work in practice.
Why it matters
Whether independent experts can freely share their findings on AI model safety and risks sits at the center of the industry's transparency debate. Limiting outside evaluation directly affects how much the public and regulators can learn about these models.
