OpenAI launches GPT-5.5-Cyber: a new era of cybersecurity amid Anthropic's restrictions
On June 22, OpenAI officially unveiled the full version of GPT-5.5-Cyber, a specialized artificial intelligence model designed to search for, verify, and eliminate vulnerabilities. This launch came at a strategically important moment, immediately after the U.S. government imposed export restrictions on Anthropic models.
Unlike mass-market products, GPT-5.5-Cyber will not be available to the general public. The model is part of the Daybreak program — OpenAI's cybersecurity ecosystem, which also includes Codex Security and partnerships with leading security companies. Access will only be granted to verified professionals working within authorized defensive scenarios.
Window of opportunity amid Anthropic restrictions
On June 9, Anthropic opened access to two versions of the Claude family: Fable 5 and Mythos 5. However, on June 12, after receiving a directive from the U.S. government under export controls, the developer was forced to disable these models. This led to serious consequences: Legion filed a lawsuit against the U.S. government, claiming that losing access to Anthropic models disrupted its tools for preparing legal documents.
OpenAI chose a different path. The company coordinated checks with federal agencies in advance and released GPT-5.5-Cyber only to verified users. This allowed it to avoid regulatory risks while simultaneously offering the market a powerful tool for cybersecurity.
Technical specifications and performance
According to OpenAI, GPT-5.5-Cyber demonstrates impressive results on specialized benchmarks. On CyberGym, the model achieved 85.6% compared to 81.8% for standard GPT-5.5. On ExploitGym, the result was 39.5% versus 25.95%, and on SEC-bench Pro — 69.8% versus 63.1%.
Interestingly, GPT-5.5-Cyber even outperformed the banned Anthropic Mythos 5 on CyberGym: 85.6% versus 83.8%. However, tests by the UK AI Security Institute painted a more complex picture. In a simulation of a corporate attack called The Last Ones, GPT-5.5-Cyber completed the scenario in 2 out of 10 attempts, while Mythos Preview did so in 3 out of 10. Later, AISI updated the data: a new version of Mythos Preview achieved 6 out of 10 attempts, while GPT-5.5 only managed 3 out of 10.
Access strategy and partnerships
OpenAI has clearly stated that GPT-5.5-Cyber is intended exclusively for verified users. For most clients, the company recommends standard GPT-5.5 with Trusted Access for Cyber and Codex Security. As part of the Daybreak initiative, the Cyber Partner Program has also been launched, including giants such as Akamai, Check Point, Cisco, CrowdStrike, IBM, Palo Alto Networks, and others.
Additionally, OpenAI announced the Patch the Planet program for open-source projects, created in collaboration with Trail of Bits and HackerOne. Among the first participants are cURL, Go, Python, Sigstore, and pyca/cryptography.
Earlier, cybersecurity agencies from the Five Eyes countries warned that advanced AI models could shift the balance of power in cyberspace within months. Artificial intelligence lowers the barrier to entry for attackers and accelerates attacks, making defense critically important.
My expert conclusion: The launch of GPT-5.5-Cyber is not just a technological step, but a strategic maneuver by OpenAI amid tightening regulations. The company shows that it is possible to develop powerful AI tools without conflicting with government agencies. However, the performance gap with Anthropic models in some scenarios suggests that the race in AI cybersecurity is just beginning. Investors and professionals should closely monitor the development of the Daybreak program — it could become a new industry standard.