OpenAI ships GPT-5.5-Cyber to rival Anthropic's Mythos

OpenAI launched GPT-5.5-Cyber in limited preview through its Trusted Access for Cyber (TAC) program for vetted defenders, supporting vulnerability identification, patch validation, and malware analysis. It directly targets Anthropic's Mythos cyber-specialist model. UK AISI separately found GPT-5.5 hits 71.4% on expert cyber tasks, but red-teamers found a universal jailbreak within 6 hours of release.