OpenAI's GPT-5.5 Matches Claude Mythos in Cyberattack Capabilities: AI Security Institute

In brief GPT-5.5 can autonomously execute sophisticated cyberattacks, completing a 32-step corporate network simulation […]