AI lets small actors run state-level hacking campaigns, Anthropic report finds

Anthropic published a report documenting misuse of its Claude models (Dec 2025–Aug 2026) that enabled a range of serious threats — Russian-aligned espionage campaigns (including a toolkit and automated evasion), Chinese-linked automated vulnerability research producing possible zero-days, ShinyHunters-affiliated cloud token dumps and rapid cloud takeovers, and large-scale model distillation by several Chinese labs that harvested model outputs; Anthropic said it disrupted activity, strengthened safeguards, and shared intelligence with partners.