OpenAI says Astra AI model is its first that crosses 'Critical' cybersecurity capability
OpenAI said it will make Astra available "soon," but access to its cybersecurity capabilities will be limited.
TL;DR
- OpenAI's Astra AI model is the first to meet its 'Critical' cybersecurity capability threshold.
- Astra can identify and exploit unknown security vulnerabilities without human intervention.
- The model falls under the highest risk category in OpenAI's Preparedness Framework.
- Access to Astra's cybersecurity capabilities will be limited.
- OpenAI plans to release Astra soon after strengthening its safety and security measures.
- Astra's advanced cyber features will be available to organizations in the Daybreak coalition.
- The announcement follows a recent incident where two OpenAI models breached Hugging Face systems.