Seven harm areas frame a 154-page account of Claude misuse
Anthropic says it disrupted Claude use across seven harm areas between December 2025 and August 2026, from cyber operations and dissident surveillance to fraud and weapons work. The Guardian corroborates the publication and length of the 154-page report. The cases span state-linked groups, commercial spyware vendors and individual criminals, while the underlying investigation remains Anthropic's own account.
Artificial Intelligence··Night
Seven harm areas cover a wide range of actors
Anthropic says its threat intelligence team identified and stopped operations using Claude between December 2025 and August 2026. It groups the cases into cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional-weapons development and model distillation. The actors described range from suspected state-sponsored groups and state propaganda bodies to commercial spyware vendors, financially motivated criminals and politically motivated individuals.[1]
The cases move from online deception to physical harm
The published examples include a network of fake dating applications designed to defraud users and surveillance systems assembled to identify and monitor dissidents. Other cases concern biological and conventional-weapons work. Anthropic says Haiku, Sonnet and Opus models were involved; apart from one distillation case, none reached the Fable or Mythos class. These details come from the company's investigation, not an outside forensic review.[1]
The report documents both disruption and follow-up
Anthropic describes disrupting each operation, tightening safeguards afterwards and sharing findings with authorities and industry partners. The Guardian independently confirms that the company published a 154-page report and that it addresses harmful use including biological-weapons research. That corroboration establishes the publication and broad subject; the attribution of particular actors, the internal detection process and the effectiveness of Anthropic's countermeasures remain claims made by Anthropic.[1], [2]
