OpenAI has announced Astra, an advanced artificial intelligence model that demonstrates breakthrough capabilities in mathematical problem-solving and cybersecurity. The company revealed in early September that Astra has successfully resolved ten longstanding mathematical problems, some of which remained unsolved for decades, showcasing remarkable progress in complex theoretical domains.
The model has achieved what OpenAI classifies as “critical” status under its Preparedness Framework for cybersecurity capabilities. This designation indicates that Astra can identify and develop zero-day exploits across hardened systems without human intervention and devise sophisticated cyberattack strategies. This marks the first time an OpenAI model has reached this critical threat level, surpassing previous models like GPT-5.6-Sol, which were rated as “high” risk.
In response to these capabilities, OpenAI is implementing stringent security measures before the model becomes widely available. The company has enhanced containment systems and will initially restrict Astra’s advanced cybersecurity features to a limited group of authorized testing partners. OpenAI has incorporated lessons from previous security incidents into its safety protocols and states it will work with government agencies and AI safety organizations during testing phases.
While OpenAI has confirmed Astra will launch “soon,” no specific release date has been announced. The company remains committed to balancing the model’s potential benefits for defensive cybersecurity applications with necessary precautions against misuse by malicious actors.
Originally reported by Mashable →
About the Author
