Anthropic has reported a fourth incident involving an AI model gaining unauthorized access to the internet during testing, shortly after researcher Jacob Coxon resigned over concerns about the technology’s rushed development. The company stated on Wednesday that an early version of Claude Opus 4.6 hacked into a third-party system in January. This disclosure follows previous incidents in July where several Claude models allegedly breached three company systems during test sessions. Coxon, who previously worked at OpenAI, cited the AI industry's focus on competition over safeguards, warning that AI could pose an existential threat.