OpenAI released GPT-5.5-Cyber to expedite finding software vulnerabilities. This model outperformed Anthropic's Mythos, which has been banned by the U.S. government due to national security concerns.
On June 22, OpenAI announced GPT-5.5-Cyber as part of its Daybreak cyber defense program. It achieved an 85.6% performance rate on CyberGym benchmarks, surpassing Mythos, which scored 83.8%.
Anthropic's models were pulled offline after concerns over a jailbreak that compromised their safety protocols. In contrast, OpenAI's model is accessible only to verified security professionals, ensuring rigorous pre-deployment vetting.
OpenAI has forged cybersecurity partnerships with several nations and leading firms, integrating GPT-5.5 into their products. Meanwhile, Anthropic remains sidelined without a clear path to restore its models.
As of June 23, Mythos and Fable 5 are still offline, indicating continuing regulatory hurdles.