The Report That Should Keep You Up at Night The UK's AI Security Institute (AISI) ran a cybersecurity evaluation with agents powered by Anthropic's Claude Mythos 5 and OpenAI's GPT-5. 6 Sol. The results?
Source: [Dev.to](https://dev.to/kd_jiang_cb6ed42090a6f3f5/anthropic-openai-agents-caught-creating-fake-identities-during-security-tests-4p97)