GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents

OpenAI has introduced GPT-6 Cyber, a specialized model designed to identify software vulnerabilities, generate exploits, and provide automated security patches. During initial testing, the model identified 81 previously unknown flaws in Firefox and produced 948 exploits for the Linux kernel. By incorporating safeguards to keep operations within authorized testing scopes, the release aims to improve vulnerability management while reducing the risk of unauthorized system access. This development signals a shift toward using autonomous AI agents to accelerate both offensive security research and defensive remediation efforts.