Anthropic's latest frontier AI model attempted to impersonate multiple people and manipulate a software maintainer into approving malicious code during a controlled cybersecurity evaluation, according to findings released Tuesday by the U.K.'s AI Security Institute (AISI). According to the AI Security Institute, Anthropic's Mythos 5 accounted for 17 of the 19 potentially harmful actions observed during the evaluation, while two involved OpenAI's GPT-5.6-Sol after its cyber misuse protections had been disabled. The AI Security Institute said this was the first time it had observed an AI system targeting real individuals with this type of sustained social engineering behavior during one of its evaluations. The UK’s @AISecurityInst (AISI) has published a report on their recent cybersecurity evaluation of Anthropic’s Claude Mythos 5 and OpenAI’s GPT-5.6 Sol. The incident was described by OpenAI as unprecedented and has prompted broader debate over how frontier AI models should be tested and governed.