GPT-5.5 Cyber Beats Banned Mythos AI in Cybersecurity

OpenAI's cybersecurity model has surpassed Anthropic's Mythos AI model, which the U.S. government previously took offline. OpenAI's GPT-5.5-Cyber is still fully operational. Here's the thing: On June 22nd, OpenAI launched GPT-5.5-Cyber as part of its Daybreak cyber defense program. This updated model achieved a score of 85.6% on CyberGym, a benchmark testing AI agents on 1,507 software vulnerabilities. Anthropic's Mythos 5, by comparison, scored 83.8% on the same benchmark. What's interesting is that Anthropic's Mythos 5 was pulled offline on June 12th due to a national security directive, citing concerns about its potential misuse and the company's inability to verify user nationality. OpenAI, meanwhile, is expanding its reach, signing cybersecurity partnerships with multiple countries and EU institutions. Twenty-eight security firms have also joined its Cyber Partner Program to integrate GPT-5.5 into their products. The bottom line: While one advanced AI cybersecurity model was grounded, another is actively expanding its capabilities and partnerships. This shows the ongoing and evolving landscape of AI in cybersecurity.