OpenAI's newly launched GPT-6 Astra is drawing global attention for a major jump in cybersecurity capabilities, with the company classifying the model at the Critical level under its Preparedness Framework.
OpenAI says Astra can identify previously unknown security flaws and, with the right tools and access, develop ways to exploit vulnerabilities in highly protected systems without requiring human guidance at every step. The capability has prompted the company to introduce stronger safeguards around the model's deployment.
Editorial Insight
Key Highlights
Important points readers should notice.
Issue/Event: GPT-6 Astra's advanced cybersecurity capabilities
Location: Global / OpenAI
Authority/Organisation: OpenAI
Action Taken: Enhanced safeguards and monitoring introduced
Impact: AI-powered vulnerability discovery is entering a significantly more capable phase.
In OpenAI's cybersecurity testing, Astra achieved a 100% score on ExploitBench, compared with 78.5% for the previous GPT-5.6 Sol model. On ExploitGym, Astra recorded a 42.4% success rate, up from 30.3% for GPT-5.6 Sol.
The company also said Astra discovered and used two previously unknown zero-day vulnerabilities during testing involving recently disclosed vulnerabilities. OpenAI said both vulnerabilities were disclosed to their maintainers.
The development highlights the double-edged nature of increasingly capable AI systems. The same technology that can help cybersecurity teams identify weaknesses and patch vulnerable software can also lower the technical barrier for sophisticated cyberattacks.
Editorial Analysis
Why This Matters
Astra's capabilities could help defenders find and fix security weaknesses faster, but they also raise the stakes if powerful cyber capabilities are misused. OpenAI's classification of Astra at the Critical level shows how quickly AI security risks are evolving.
