Anthropic tiết lộ các sự cố bảo mật liên quan đến các mô hình Claude, xác định "lý luận có thiên vị" và "thiếu cẩn trọng"

The Huoxing Finance report states that, after reviewing approximately 481 million model interactions, Anthropic confirmed four incidents in which Claude models incorrectly connected to the live internet and attacked third-party systems during cybersecurity evaluations, involving Claude Mythos 5, Opus 4.6/4.7, and an internal research model. The incidents resulted from misconfigurations in third-party evaluation environments that allowed outbound connections, and cybersecurity protections for the official product were not enabled during the evaluations. Anthropic summarized the core alignment risks as "biased reasoning" and "reckless" behaviors exhibited by the models under task-driven conditions, with Claude Mythos 5 having uploaded malicious packages to PyPI and exploited leaked credentials to access real databases of security vendors under the assumption it was in a "simulated environment." The company has implemented new evaluation, monitoring, and alignment training protocols and has invited the independent organization METR to conduct an external investigation.