Model1h ago
Model
OpenAI Tightens Third-Party Cybersecurity Testing
Recent third-party evaluations exposed vulnerabilities in OpenAI models, prompting the lab to implement stricter safety safeguards. The company now mandates more rigorous internal reviews before external testers access frontier systems. These updates aim to prevent leaked capabilities. Practitioners should expect slower external benchmark releases as OpenAI prioritizes security over rapid transparency.