Anthropic and OpenAI reportedly seek control over evaluator publications, raising trade secret concerns

Anthropic official brand assets (anthropic.com) Anthropic and OpenAI have reportedly taken steps to exert significant control over what external evaluators can publish about their model-safety reviews. The move is said to stem from concerns that these findings could inadvertently reveal trade secrets. This development highlights a tension between the need for independent safety validation and the protection of proprietary information. Both Anthropic, known for its Claude model, and OpenAI, the company behind ChatGPT, are involved in ongoing negotiations around these issues. The situation underscores the complex balance companies must maintain as they navigate the competitive landscape of AI advancements. Key Takeaways Reports suggest Anthropic and OpenAI are seeking to manage the publication of external evaluations, which could indicate a strategic effort to protect trade secrets. Market pricing shows an increased perception of potential financial distress for OpenAI, which may be consistent with concerns over operational challenges. Current market odds for OpenAI announcing bankruptcy by the end of 2027 indicate a noticeable increase, reflecting participants’ interpretation of rec...