On June 22, OpenAI released the full version of GPT-5.5-Cyber, a specialized model designed for identifying, verifying, and fixing vulnerabilities. This launch occurred against the backdrop of restrictions surrounding Anthropic.

According to the announcement, GPT-5.5-Cyber is part of the Daybreak program and will not be available to the general public. The model is intended for vetted cybersecurity professionals who require enhanced capabilities for authorized protective work.

How OpenAI Found Its Opportunity

On June 9, Anthropic opened access to two versions of its Claude model family. The company described Fable 5 as a Mythos-class solution, but safe for general use. Claude Mythos 5 is a "private" base model with relaxed restrictions in certain areas. However, by June 12, the developer disabled them following a directive from the U.S. government under export controls.

This decision caused issues not only for regular users. On June 23, Legion filed a lawsuit against the U.S. government due to the directive, as reported by Reuters. The organization is based in San Jose, but its development team is located in Canada. According to the agency, Legion stated that losing access to Anthropic's models disrupted its tools for preparing legal documents and managing cases.

In this context, OpenAI chose a different path: the company clarified that it had coordinated checks with U.S. federal agencies beforehand and then opened GPT-5.5-Cyber only to vetted users.

Capabilities of GPT-5.5-Cyber

According to OpenAI, GPT-5.5-Cyber achieved 85.6% on CyberGym compared to 81.8% for the standard GPT-5.5. This metric focuses on reproducing known vulnerabilities in controlled software environments and does not cover the full range of real attacks and defensive scenarios.

Source: OpenAI.

OpenAI's blog also highlighted results from other tests. On ExploitGym, the model scored 39.5% versus 25.95% for the standard GPT-5.5. On SEC-bench Pro, it scored 69.8% compared to 63.1%. The first test evaluates the ability to turn a known vulnerability into a working exploit, while the second assesses long-term vulnerability discovery and proof-of-concept creation in complex software targets.

Source: OpenAI.

OpenAI emphasized that the goal of Daybreak is not just to find more vulnerabilities but to accelerate their remediation. Since March, the cloud version of Codex Security has scanned over 30 million commits across more than 30,000 codebases, with human reviewers marking over 70,000 findings as fixed.

According to Decrypt, GPT-5.5-Cyber outperformed Anthropic Mythos 5 on CyberGym: 85.6% versus 83.8%. However, other public assessments present a more complex picture. On April 30, the UK AI Security Institute reported that GPT-5.5 completed a 32-step simulation of the corporate attack The Last Ones from start to finish in 2 out of 10 attempts. Mythos Preview succeeded in 3 out of 10 attempts for this scenario.

Later, AISI updated its data: the new version of Mythos Preview completed The Last Ones in 6 out of 10 attempts and successfully finished the second scenario, Cooling Tower, in 3 out of 10 attempts. In this update, GPT-5.5 completed The Last Ones in 3 out of 10 attempts.

How OpenAI Restricts Access

GPT-5.5-Cyber is not intended for general access. OpenAI clarified that the model is designed for vetted users who require stronger cyber capabilities and the model's permissive behavior in authorized scenarios. For most clients, the company still recommends GPT-5.5 with Trusted Access for Cyber and Codex Security.

OpenAI also launched the Daybreak Cyber Partner Program. In its blog, the company listed participants including Akamai, Check Point, Cisco, CrowdStrike, IBM, Palo Alto Networks, Proofpoint, SentinelOne, Wiz, Zscaler, and other security firms.

Source: OpenAI.

Another initiative is Patch the Planet, aimed at open-source projects. This program was created in collaboration with Trail of Bits, involving HackerOne, Calif, researchers, and maintainers. Among the first participants are cURL, Go, Python, Sigstore, and pyca/cryptography.

Previously, cybersecurity agencies from the U.S., U.K., Canada, Australia, and New Zealand stated that advanced AI models could change offensive and defensive capabilities in cyberspace within months rather than years. The report's authors believe that artificial intelligence lowers the entry threshold for attackers, accelerates attacks, and shortens the window between vulnerability discovery and exploitation.

In June, the Anthropic team published a blog post on Claude titled Zero Trust for AI agents, discussing the safe deployment of autonomous AI agents in corporate environments. The document outlines key risks of agent systems and approaches to business cybersecurity.