OpenAI is rolling out GPT-5.5-Cyber to a restricted pool of "critical cyber defenders" starting this week, CEO Sam Altman said on X, but the company is deliberately taking a wider distribution approach than rival Anthropic. The model is a specialized version of GPT-5.5, OpenAI's flagship model released earlier this year. It will not reach the general public.
Altman said the limited rollout begins "in the next few days," with the company working alongside government and industry partners to determine who qualifies.
"We believe the better path is responsible, trusted access for defenders so they can move faster than adversaries can adapt," OpenAI told Nextgov/FCW in a statement. The move puts OpenAI on a collision course with Anthropic's philosophy. Anthropic released its own frontier cybersecurity model, Claude Mythos, earlier this month to roughly 50 organizations in a tightly controlled program. The company has stated Mythos will never be made publicly available. OpenAI plans to distribute GPT-5.5-Cyber more broadly through its Trusted Access for Cyber (TAC) program to government entities, critical infrastructure operators, security vendors, cloud platforms, and financial institutions. The model builds on GPT-5.4-Cyber, introduced in mid-April alongside $10 million in API grants for vetted security organizations.
Anthropic's Mythos demonstrated the ability to autonomously discover thousands of zero-day vulnerabilities, including constructing a full browser exploit and a FreeBSD remote code execution exploit through complex vulnerability chaining. OpenAI has not published technical benchmarks comparing GPT-5.5-Cyber to Mythos, and the company has not released any technical specifications for the new model. The White House has taken a keen interest in Mythos' rollout, and has recently opposed plans to expand access to the model further, according to The Wall Street Journal. Unnamed White House officials cited cybersecurity concerns and worries that increased demand would hamper the government's ability to use the system.
Anthropic is still dealing with the fallout from a Pentagon supply chain risk designation after the company refused to allow its products to be used in autonomous weaponry or surveillance of Americans. A judge issued an injunction on that designation, though the government plans to appeal. At least one agency, the State Department, has already switched from using Claude Sonnet 4.5 to OpenAI's GPT-4.1 for its internal AI system.
Alongside the model release, OpenAI published a Cybersecurity Action Plan structured around five pillars: democratizing access to cyber defense tools, coordinating with government and industry, enhancing safeguards around advanced capabilities, ensuring deployment visibility, and enabling user self-protection. The plan signals OpenAI intends to push for broader distribution of defensive AI capabilities even as the industry trend moves toward tighter restrictions.
Altman said the company will "work with the entire ecosystem and the government to figure out trusted access for Cyber," adding that the goal is to secure critical infrastructure. OpenAI has already updated its contracts with Microsoft and is engaging with Amazon Web Services to offer its agentic tools across multiple cloud providers.
