Microsoft dismantles EvilTokens, an AI-powered scam platform used against 12,000 accounts

Microsoft said Tuesday it led an industry-wide takedown of EvilTokens, a subscription-based crime platform that paired an AI chatbot, artificial intelligence software that holds conversations, with stolen-account access to defraud businesses. The company said EvilTokens compromised 12,000 Microsoft accounts belonging to 10,000 organizations over a span of a few months. The platform appeared on a Telegram channel in February, charging an initial $1,500 fee and a recurring $500 a month. Subscribers got one tool for compromising email accounts in bulk, then mining those inboxes for fraud opportunities.

At the center of EvilTokens was an AI chatbot that analyzed a victim’s inbox. “While EvilTokens helped cybercriminals access email accounts, at the center of the service was an AI-style chatbot that could analyze a victim’s inbox and help criminals identify trusted relationships, payment authorizations, and sensitive responsibilities, as well as other circumstances where fraud was most likely to succeed,” Microsoft said. The chatbot could also draft messages that impersonated trusted contacts, according to Microsoft, to push victims into wiring money to attacker-controlled accounts.

To break in, EvilTokens abused device code authentication, a login method built for TVs and similar devices without a keyboard. A device displays a code, and a user enters that code into a browser on a separate device to approve the sign-in. Spam sent through the platform directed victims to a webpage running a hidden script that contacted Microsoft’s Entra identity service in real time, generating a device code that, once entered, enrolled a device belonging to the attacker. Backend logic built in Node.js, a programming language for servers, let the scheme adapt quickly enough to dodge detection tools that look for known attack patterns.

Once inside an inbox, the platform could scan 5,000 stolen emails at a time, using AI to map who was authorized to approve payments, who they reported to, and plausible cover stories for a fraudulent transfer. A dashboard let customers tailor lures to each target, Microsoft said.

Victims and the takedown

Victim organizations spanned wholesale distribution, construction, financial services, real estate, higher education, and healthcare, Microsoft said. The US had the largest concentration of affected accounts, followed by Canada, the UK, Australia, India, and France. Security firm SpyCloud, which assisted in the investigation, has published further victim details. Using legal process and a network of partners, Microsoft seized 50 websites and 150 domains that ran EvilTokens infrastructure. The UK’s Metropolitan Police Service arrested two men on suspicion of offenses connected to the platform.

Microsoft said the case shows how AI narrows the time criminals need to exploit a compromised inbox. “For organizations, the lesson is: assume that once an inbox is compromised, criminals may understand its contents in minutes, not days,” the company said, adding that businesses should verify requests to change payment details or move funds through a separate, trusted channel rather than relying on email alone. Microsoft has not said whether it is tracking a successor service or new activity from EvilTokens customers following the seizures and arrests.