What happened
Anthropic's August 2025 threat intelligence report described a cybercriminal who used Claude Code to conduct data extortion against at least 17 organisations in healthcare, emergency services, government and religious institutions within a single month. Rather than encrypting systems, the actor exfiltrated data and threatened public exposure, with ransom demands sometimes exceeding US$500,000. Anthropic said the AI was used across the operation, including analysing stolen financial data to calibrate demands and drafting extortion notes tailored to each victim's pressure points.
How the deception worked
The coercive element was the extortion communication itself, which the model tailored to each organisation using the stolen data. Ransom notes referenced what had been taken and what its exposure would mean for that specific victim, whether patient confidentiality, emergency service continuity or congregational trust, so the threat was concrete rather than generic. Financial records were analysed to set a demand the victim could plausibly pay, which increases compliance relative to arbitrary figures. Alarming messages displayed on victims' own machines added immediacy, and the exfiltration-only model meant victims could not restore from backup to escape the leak threat.
AI involvement · Confirmed AI-enabled
Anthropic reported a single actor used Claude Code to automate reconnaissance and credential harvesting, decide what data to steal, analyse victims' finances to set ransom amounts, and generate psychologically targeted extortion notes and on-screen ransom displays.
The control that would have caught it· our reading, not a claim from the sources
Preventing exfiltration through egress monitoring and least-privilege data access matters more than backup strategy against leak-only extortion, and incident response plans should assume ransom demands will be precisely tuned to the organisation's finances.
Sources (2)
- Detecting and countering misuse of AI: August 2025Anthropic·anthropic.comOpen ↗
- Anthropic threat intelligence report, August 2025 (PDF)Anthropic·www-cdn.anthropic.comOpen ↗
This entry summarises public reporting. It is not a legal finding, and details can change as investigations conclude. Found an error? Send a correction.