Microsoft Takes Down EvilTokens, an AI Service Built for Email Fraud

Microsoft and a group of technology and law-enforcement partners have disrupted EvilTokens, an alleged cybercrime service that used an AI-style chatbot to turn compromised email inboxes into tailored fraud plans. The action seized 50 operating websites and disabled more than 150 associated domains under a U.S. court order.

Microsoft’s account describes a service built to automate the slow, investigative part of business-email fraud. Rather than merely helping an attacker draft a persuasive message, EvilTokens could analyze a victim’s mailbox, map organizational roles, surface payment conversations and identify trusted relationships. It could then recommend people to impersonate and targets for fraud.

From a sign-in trick to a fraud roadmap

The alleged entry point was a device-code phishing attack. Victims were tricked into entering an authentication code on Microsoft’s legitimate sign-in page, giving criminals access without handing over a password. Microsoft said that access could remain after a password reset if associated sessions and tokens were not also revoked.

Once inside, the service’s alleged value was speed and structure. A criminal who had obtained access no longer needed to manually sift through thousands of messages to find decision-makers, invoices or payment procedures. Microsoft said preset prompts could locate wire-transfer discussions, identify people who move money and suggest the best contacts to impersonate.

A subscription product for a criminal workflow

Microsoft said EvilTokens was sold through Telegram for a $1,500 initiation fee and a recurring $500 subscription. Investigators also found evidence that large parts of the platform had been built with AI assistance and that it drew on capabilities from multiple AI models. The service paired those tools with dashboards, support and fraud-focused features in what Microsoft characterized as a commercially run operation.

That packaging is central to the risk Microsoft identifies: expertise once spread across identity attacks, cloud systems, social engineering and financial fraud was presented through a ready-made interface. The company said the platform affected organizations across sectors including financial services, real estate, higher education, healthcare, construction and wholesale distribution.

The takedown paired civil action with arrests

The U.S. District Court for the Eastern District of Virginia authorized the disruption. Microsoft and Health-ISAC worked with Cloudflare, Coinbase, OpenAI, Railway, SpyCloud, The Shadowserver Foundation and TRM Labs against parts of the service, while Microsoft said it notified affected customers and helped remediate compromised accounts.

In the United Kingdom, Metropolitan Police arrested two men, ages 32 and 38, on September 11 on suspicion of offenses connected with the alleged EvilTokens operation. Both were released on police bail subject to conditions while the investigation continues. Microsoft called the case its Digital Crimes Unit’s first court-authorized action against an end-to-end AI-enabled cybercrime service.

The remaining problem is the inbox itself

Removing a service’s infrastructure does not remove the underlying playbook. Microsoft’s practical warning is that a compromised inbox may now be understood in minutes rather than days. Its advice is to use strong identity protections and independently verify requests to change payment information, redirect funds or approve unusual transactions through a trusted second channel.