Anthropic’s Claude Misuse and Alignment Saga Expands Across Cyber, Bio, Surveillance, and Model Provenance

Anthropic’s expanded assessment and related reporting describe hacking incidents, fake developer personas, attempted supply-chain attacks, biological-threat assistance, and Claude use in surveillance and malicious state-linked operations. Anthropic also alleges that Moonshot routed some Kimi requests to Claude for training, while AISI access to Mythos 5.1 remains uncertain. The story is at peak activity as independent-evaluation access, adaptive attacks, and safeguard robustness remain unresolved.