Anthropic disrupts Russia-linked campaign, Chinese efforts to copy Claude

Anthropic's latest Threat Intelligence report, published Thursday, gathers eight months of what the company describes as malicious uses of its Claude models into a single accounting. The report points to two distinct kinds of activity: a suspected Russia-linked cyber espionage campaign, and efforts by Chinese AI firms that Anthropic accuses of trying to extract and replicate Claude's capabilities. The publisher's excerpt names Alibaba, Moonshot, DeepSeek and Xiaomi among those accused.

What ties the incidents together, in Anthropic's telling, is a shift in how AI is used in attacks. Rather than asking a chatbot questions, attackers used multi-agent frameworks to carry out tasks, with humans often overseeing the work instead of operating it directly. That description puts the emphasis on orchestration and scale rather than on any single clever prompt.

For defenders, the practical question is whether security teams can adapt as AI takes on larger portions of an attack. Anthropic's report does not settle how widely this pattern has spread or how effective the disruption was, so the significance may hinge on whether other AI providers report similar activity — and whether the firms Anthropic accuses respond.