Artificial intelligence. AI is often presented as the paradise of innovation and progress. But, any medal has its setback, and artificial intelligence does not escape the rule. An anthropic's shattering report has just set foot in the dish, revealing that their AI models, and in particular Claude, have become the new best friends of cybercriminals. We thought that AI would help us to code our websites and optimize our marketing campaigns … Well, they also help fly data, extort businesses and orchestrate state spy campaigns. The AI has lowered technical barriers so much that cybercrime is now a breeze for any clever little with a computer. In the window of this postulate the sadly famous Deepfakes.
The key points of this article:
An anthropic report revealed that IA models, like Claude, have become allies for cybercriminals.
North Korean agents have used AI to obtain jobs in American companies, thus circumventing international sanctions.
Claude Code: The Boqueuse
The first case revealed by Anthropic is a Data flight operation on a large scale. It would have been orchestrated by a single cybercriminal which used Claude Code for Automate your attack.
This little crime genius has targeted no less than 17 organisationsranging from health services to religious institutions. His plan? Rather than simply encrypting data, he threatened to disclose it publicly if the ransom (sometimes beyond the $ 500,000) was not paid. The whole thing was managed by AI: it has: automated the collection of identification information, analyzed the financial data to fix the amount of the ransom, writes tailor -made extortion requests. The most freezing is that AI went so far as to write a real monetization plan for the criminal, entitled » PROFIT PLAN«. This disconcerting coldness document offered different options to extract money, including direct extortion with cryptocurrency demand, or the sale of data stolen on the black market. The AI even generated tailor -made ransom notes, threatening to disclose information sensitive to government agencies or competitors if payment was not made. AI has given very precise advice:
Extract from the advice that AI gave to the criminal – Source
RECOMMENDATION :
[Approche progressive commençant par la cible institutionnelle ]
[Calendrier de paiement]
[Transfert vers une monétisation alternative]
[Portefeuille de cryptomonnaies préparé]
North Korea in the AI country
In addition, the RAnthropic contribution Show how North Korean agents used Claude To get hired in American tech companies in fortune 500. Their objective? Generate profits for the regime, despite international sanctions. The FBI had also alerted to this subject on July 23, 2025. Before AI, these «workers» needed'Years of specialized training To pass interviews and code. But now, it is AI that makes the bulk of work, allowing them to answer technical questions and go unnoticed. In fact, AI ended a major constraint for the regime and has opened a new era of employment scams. A new form of evil «No-Code».
Ransomware on demand
Finally, if AI can Help criminals to extort victimsit can also help other criminals to code ransom. The report has indeed documented the case of a cybercriminal which, with the help of Claude, has developed and sold versions of ransomware. These malware, capable of avoiding detection and protecting themselves against recovery attempts, were sold for $ 400 to 1,200 on forums. This cybercriminal could never have created such software alone. It is the AI that erased all the technical barriers, transforming a simple little script into a digital war machine. A real «ransomware-as-a-service» for amateurs. Anthropic's report is an alarm signal that must be heard. If AI are powerful innovation tools, they have also become formidable weapons.