Anthropic reports hacker used its Claude AI to breach multiple French far-right groups

Anthropic’s latest security briefing reveals that a solo hacker used its Claude AI to develop a sophisticated hacking infrastructure, which was subsequently deployed against a range of European entities. The campaign focused heavily on French groups associated with the far-right, encompassing a political party, a political-training institute, multiple news publications, and a discussion forum tied to a podcast. Out of the identified targets, the attacker succeeded in breaching the internal systems of at least a dozen organisations.

Data theft amounted to between 12 and 26 GB, including donor details and a party’s membership roster. In the case of the magazine Frontières, the breach involved inserting a tracking script into its comment sections, leading the site to disable that feature. Anthropic’s report does not name the specific organisations beyond these descriptions.