
Security modelFinds exploitable vulnerabilities and develops fixes
GPT-6 Cyber develops over five times as many Linux exploits
GPT-6 Cyber, OpenAI’s cybersecurity model, developed working exploits for 948 Linux vulnerabilities versus 166 for its earlier GPT-5.6 Sol, OpenAI reports. It is initially available through the API and Codex, OpenAI’s coding tool, to selected approved Daybreak Red customers.
Sources
Official releaseGPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agentsWritten by AI from the sources