Anthropic says AI now lets lone operators run state-level hacking campaigns
Individual hackers are now sustaining campaigns that would have needed many skilled operators a year ago, because artificial intelligence has absorbed the labor that used to set state-backed teams apart. That’s according to AI model developer Anthropic PBC, which today published its latest threat intelligence report, covering Claude misuse it disrupted between December and August. […] The post…
Anthropic, the AI model developer, published a threat intelligence report on AI misuse between December and August. The report reveals that AI has reduced the need for large teams of skilled hackers, allowing lone operators to carry out state-level hacking campaigns. Seven Chinese AI labs were accused of illicitly distilling Claude's capabilities.
The report covers seven harm areas, including breaches conducted by suspected state-sponsored groups, financially motivated criminals, and commercial spyware vendors. They used Claude Haiku, Sonnet, and Opus models. Claude enabled breaches completed in two to three hours and handling dozens of victims simultaneously. The report highlights a case where Chinese-speaking group GTG-10007 compromised Azure Active Directory and stole over 2,100 token sets.
Another group, GTG-50020, used Claude to access production application programming interface keys for several AI model providers. Distillation attacks were carried out by an Alibaba Group-affiliated actor, forwarding over 300,000 requests to Claude Opus 4.6 and 4.7 daily. The report also mentions the disruption of six weapons development cases in China, Russia, and Yemen.
Anthropic implemented measures to prevent AI misuse, such as summarizing Claude's internal reasoning and restricting access for accounts operating from unsupported countries.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
