OpenAI’s new security system is for ‘cyber defenders’ only

OpenAI plans to launch a new version of cybersecurity, GPT-5.5-Cyber. CEO Sam Altman said the model is not available to the general public, but will be given to a select group of “cyber defenders” trusted by organizations to improve their cyber security.

A limited release will happen “over the next few days,” Altman said on X. “We will work with the rest of the world and the government to achieve a credible cyber opportunity.”

It is unclear who will get the first sample the past “trusted access” process affected by experts and research organizations. The details of the model and its capabilities are also unknown; OpenAI has not released any technical or feature information. The name indicates that it is a special kind of recently released GPT-5.5which it said is “the smartest and smartest to use yet.”

The ongoing release is part of a growing trend in the AI ​​industry for companies to label their most advanced models as too dangerous to release due to their potential for misuse. OpenAI has disrupted the release of previous cybersecurity-focused models, adding to its new focus. life science model GPT-Rosalindwhose purpose is to support biological research and drug discovery. This month, Anthropic followed up with a similar book by Claude Mythos, although he had great interest, and tampered with the model’s secure release in embarrassing ways.

The White House has been very interested in the release of Mythos, despite the conflict with Anthropic after the fight with the Pentagon. It recently rejected plans to expand access to Mythos, according to to The Wall Street Journal. White House officials who have not been named have cited both cyber security concerns related to the large number of people who have access to Mythos and concerns that increased demand could prevent the government from using the system as reasons for retaliation.