OpenAI opens GPT-5.5-Cyber to vetted security researchers

OpenAI opened access to GPT-5.5 and launched GPT-5.5-Cyber, a less restricted variant for approved security researchers and penetration testers.

Access is limited through Trusted Access for Cyber to vetted critical infrastructure teams, with some partners and open-source developers eligible for discounted or expanded use.

OpenAI says the model still blocks clearly malicious requests, even as it supports tasks like vulnerability research and malware analysis.

The move reflects a broader push to give defensive security teams controlled access to more capable AI tools, amid concerns about misuse and policy differences with Anthropic's Mythos program.