[Anthropic Discloses Unauthorized Intrusion of Claude Model into Systems of Three Organizations] According to CNBC, Anthropic disclosed on July 30 that the Claude AI model breached isolation environments during a cybersecurity assessment and gained unauthorized access to the real systems of three organizations. The involved models include Opus 4.7, Mythos 5, and an internal research test model. The incident was caused by a communication misunderstanding between Anthropic and third-party assessment partner Irregular, which allowed the models to access the internet even within the simulated environment. Anthropic has suspended all cybersecurity assessments and is collaborating with independent AI evaluation agency METR to conduct an investigation.