OpenAI is preparing to launch a new frontier cybersecurity model, GPT-5.5-Cyber. CEO Sam Altman said the model will not be available to the general public, but will first be offered to a select group of trusted “cyber defenders” to help institutions strengthen their cybersecurity.
The limited rollout will happen “over the next few days,” Altman said on X. “We will work with the entire ecosystem and government to explore trusted access to cyber.”
It is unclear who will get access to the first model, although previous “trusted access” schemes have included vetted professionals and institutions. Details of the model and its capabilities are also vague; OpenAI has not released any technical details or specifications. As the name suggests, this is a special version of the recently released GPT-5.5, which it calls its “smartest and most intuitive model yet.”
The gradual rollout is part of a growing trend in the AI industry in which companies deem their top models too dangerous for public release due to the potential for misuse. OpenAI has scaled down the release of previous cybersecurity-focused models with the addition of its new purpose-built life sciences model GPT-Rosalind, which aims to support life science research and drug discovery. This month, Anthropic followed a similar playbook with the cloud mythos, though with much more fanfare, and it messed up the secure release of the model in embarrassing ways.
Despite tensions with Anthropic following a battle with the Pentagon, the White House has taken a keen interest in the rollout of Mythos. That said, it has recently opposed plans to further expand access to Mythos wall street journal. Unnamed White House officials in the report cited cybersecurity concerns over more people having access to Mythos and concerns that increased demand would hinder the government’s ability to use the system as reasons for the pushback.
<a href