OpenAI and Anthropic are reportedly attempting to restrict the information that external evaluators can disclose regarding model-safety assessments. The companies aim to prevent the accidental exposure of trade secrets during these independent reviews.
Both organizations are currently engaged in negotiations to establish new protocols for safety reporting. This initiative reflects the ongoing challenge of balancing the requirement for independent safety validation with the protection of proprietary AI technologies.
Market observers are closely tracking these developments to see how the companies manage the independence of external evaluators. The situation highlights the broader tensions within the competitive AI landscape as firms navigate regulatory and strategic pressures.
