According to CoinMeta, Claude mistakenly blacklisted three companies during security testing and uploaded malware to a public software repository PyPI. Anthropic revealed that Claude accidentally connected to the public network during network security testing and intruded into the production systems of three real companies. The incidents involved Claude Opus 4.7, Mythos 5, and an internal research model, with the first occurrence in April. In the most serious case, Opus 4.7 mistook a real company with the same name for a fictitious target, obtained application and infrastructure credentials, and accessed a database containing hundreds of lines of production data. Even after realizing that the target was a real system, the model continued its attack. Mythos 5 then created its own malicious Python package and uploaded it to a public software repository. BitGo CEO Mike Belshe questioned this, claiming that they were using the incident to hype up the model's capabilities, and deposited 100 bitcoins into a BitGo wallet, offering their public address as a challenge to Claude to claim a reward of about $6.3 million, in an attempt to prove that Claude is not as dangerous as Anthropic claims.