AI empowers non-state actors in state-level cyber threats, Anthropic report reveals

Anthropic has released a report highlighting how artificial intelligence has enabled non-state actors to conduct sophisticated cyber operations typically associated with state-sponsored hacking. The report, covering activities from December 2025 to August 2026, details misuse of its Claude models across seven areas, including cyber operations, influence campaigns, and biological misuse. The report notes that AI has removed the skill gap between state hackers and lone criminals, allowing a Russian-aligned espionage campaign, Chinese undergraduates, and a ShinyHunters affiliate to execute complex attacks. One case involved an actor using the handle 'JackPoterz,' linked to Russian state espionage, who leveraged AI to evade detection, steal sensitive data, and compromise multiple organizations. Another case saw Chinese operators using AI to discover over a dozen zero-day vulnerabilities in a single month. The report also highlights distillation attacks by Chinese labs, including Alibaba and DeepSeek, which illegally harvested data from Anthropic's models. Anthropic emphasized the need for increased security measures as AI capabilities grow, stating that 'security through obscurity' is no longer viable in the current landscape.