Responding to the next frontier of critical cyber capabilities
OpenAI announced that preliminary evaluations of its upcoming Astra model suggest it may achieve Critical cybersecurity capabilities, leading to enhanced security measures.
- Astra could identify and develop functional zero-day exploits in many hardened real-world critical systems without human intervention.
- OpenAI is implementing stricter security controls, including isolated testing environments and enhanced model weight protections.
- Internal activities involving Astra that don't meet strengthened security requirements are paused.
- OpenAI will work with government agencies and AI safety organizations to test Astra's capabilities.
- Previous models like GPT-5.6-Sol were assessed at High, not Critical, cyber capability threshold.
AI developers and security teams must prepare for models that can autonomously exploit critical systems, changing how they assess and mitigate cyber risks.
Sources
Read this as text
Back to the AI news