OpenAI Unleashes Cyber AI for Exploit Dev

OpenAI recently introduced GPT-5.6-Cyber, a new AI model specifically engineered for advanced cybersecurity operations, including vulnerability research and exploit development.

This specialized model, based on GPT-5.6 Sol, demonstrates a dramatically improved ability to handle complex cybersecurity tasks, with internal benchmarks showing a 95% completion rate for exploit-chain development and similar challenges. Crucially, it is designed to reduce refusals on requests that may have dual-use applications.

Access to GPT-5.6-Cyber is not widespread; it is exclusively available through OpenAI's new Daybreak Red cybersecurity program tier, which requires rigorous vetting of participating organizations. A broader tier, Daybreak Blue, offers adjusted general models for cybersecurity uses.

OpenAI has confirmed that GPT-5.6-Cyber has already contributed to identifying significant security flaws, including two zero-day vulnerabilities in Google's V8 JavaScript engine, which were subsequently disclosed and patched by Google.

The introduction of these specialized models highlights a growing trend in AI-assisted offensive security, with OpenAI positioning GPT-5.6-Cyber as a tool for approved defenders, emphasizing strict access controls and monitoring.