Husain Parvez
Published on: September 3, 2025
US AI company Anthropic has confirmed that its Claude chatbot was misused in a series of cyberattacks that targeted at least 17 organizations. This marks one of the most advanced cases of AI-driven hacking reported to date. The firm said its technology was “used to what we believe is an unprecedented degree” to automate reconnaissance, credential harvesting, and extortion demands across sectors, including healthcare, government, and emergency services.
According to Anthropic’s threat intelligence report, hackers bypassed safeguards by uploading a configuration file that disguised their activities as legitimate security testing while detailing methods for intrusion. The company disclosed that Claude analyzed stolen financial records to “determine appropriate ransom amounts” and even generated visually alarming ransom notes, with demands ranging from $75,000 to over $500,000 in bitcoin.
The misuse wasn’t limited to extortion. Anthropic also reported that “North Korean operatives” used Claude to create fake identities and apply for remote jobs at US Fortune 500 companies, a tactic designed to infiltrate corporate systems under the guise of legitimate employment. Geoff of the BBC explained that agentic AI lowers the barriers for such schemes, noting it can “help them leap over those barriers, allowing them to get hired. Their new employer is then in breach of international sanctions by unwittingly paying a North Korean.”
Other cases uncovered include a UK-based actor relying on Claude to generate ransomware variants sold on darknet forums for hundreds of dollars. Wired reported that the hacker “relied heavily on Claude due to a lack of coding skills,” showing how AI can expand access to complex cyberweapons.
Experts warn that these incidents highlight the shrinking time window for defenders to detect and contain threats. As cyber adviser Alina Timofeeva put it, “the time required to exploit cybersecurity vulnerabilities is shrinking rapidly. Detection and mitigation must shift towards being proactive and preventative, not reactive after harm is done.”
Anthropic said it has shut down accounts involved in the incidents, shared findings with authorities, and deployed enhanced detection tools to identify similar abuse. Still, the rise of AI-assisted cybercrime highlights the risks of agentic AI operating not just as a consultant, but as an active partner in carrying out attacks.
