OpenAI is gearing up to release a new cybersecurity model, GPT-5.5-Cyber, and if you were hoping to use it to automate your phishing scams, you're out of luck. CEO Sam Altman announced that the model will not be available to the general public but will instead be rolled out to a select group of trusted “cyber defenders” to help institutions shore up their defenses.
The limited rollout will happen “in the next few days,” Altman said on X, adding, “We will work with the entire ecosystem and the government to figure out trusted access for Cyber.” Who exactly gets access remains unclear, though previous “trusted access” schemes have involved vetted professionals and institutions. OpenAI has also not released any technical details about the model, though its name suggests it's a specialized version of the recently launched GPT-5.5, which the company called its “smartest and most intuitive to use model yet.”
This staggered release is part of a growing trend in the AI industry where companies brand their top models as too dangerous for public release due to potential misuse. OpenAI has done this before with previous cybersecurity models, as well as with GPT-Rosalind, a life sciences model intended to support biology research and drug discovery. Anthropic recently followed a similar playbook with Claude Mythos, though with much greater fanfare - and a bungled secure release that was embarrassing in ways that are best left unexamined.
The White House has taken a keen interest in Mythos's rollout, despite lingering tensions with Anthropic after its fight with the Pentagon. According to The Wall Street Journal, unnamed White House officials have opposed plans to expand access to Mythos further, citing both cybersecurity concerns and worries that increased demand would hamper the government's ability to use the system. Because nothing says “trust us” like the government hoarding the AI for itself.