Responding to the next frontier of critical cyber capabilities

OpenAI has determined that one of its upcoming models has reached a "Critical" level of cybersecurity capability, meaning it can identify and develop functional zero-day exploits in hardened systems or devise novel cyberattack strategies without human intervention. The company established a Preparedness Framework to guide its response as models approached dangerous capabilities in areas like cybersecurity, and this assessment triggered that protocol. In response, OpenAI has implemented stricter security controls including isolated testing environments, restricted access, enhanced encryption, and universal monitoring for risky actions, while also pausing internal activities that don't meet these new requirements and planning to work with government agencies and safety organizations on testing. This represents a significant shift from previous models, which had only reached a "High" rather than "Critical" threshold for cyber capabilities.