OpenAI rolls out limited GPT-5.5 cyber release, first to vetted security experts

OpenAI is releasing GPT-5.5 Cyber in a limited rollout to vetted cybersecurity experts first.

Access will be application-based, with broader availability planned later in coordination with U.S. officials and qualified users.

The model is meant to support defensive work like penetration testing, vulnerability detection, and malware reverse engineering.

The restricted launch reflects growing concern that powerful security AI can help defenders but also be misused offensively.