OpenAI prepares GPT-5.5-Cyber for trusted security researchers

OpenAI plans to roll out GPT-5.5-Cyber, a cybersecurity-focused version of its flagship model, to a limited group of trusted defenders.

Sam Altman said access could begin within days, but OpenAI has not disclosed technical details or the first wave of recipients.

The restricted launch continues OpenAI’s earlier controlled cyber releases and mirrors similar guarded deployments by other AI companies.

The move highlights growing efforts to limit misuse of powerful AI for cyber offense while keeping access for vetted security and government users.