Microsoft dismantles chatbot-driven EvilTokens fraud platform linked to 12,000 hacked inboxes

Microsoft said on 22 September 2026 that it had disrupted EvilTokens, a subscription-based cybercrime platform built around an AI chatbot, which had been linked to more than 12,000 compromised email inboxes at over 10,000 organisations worldwide within months of launching in February 2026, according to a post by Steven Masada, general manager of Microsoft's Digital Crimes Unit (DCU). Ars Technica confirmed the account the same day, citing Microsoft's disclosure.

How the platform worked

• EvilTokens was sold through a Telegram channel for a $1,500 sign-up fee plus a $500 recurring monthly charge, Microsoft said.
• Criminals tricked victims into completing a legitimate Microsoft device-code sign-in process, giving attackers account access without stealing passwords, according to Microsoft.
• Once inside an inbox, the AI chatbot could summarise and translate emails, surface financial conversations, map organisational roles and identify trusted relationships, including finding an organisation's "money movers" and drafting impersonation messages, according to Microsoft.
• Victim activity was concentrated in the United States, Canada, the United Kingdom, Australia, India and France, spanning wholesale distribution, construction, financial services, real estate, higher education and healthcare, Microsoft said.

The takedown

With authorisation from the US District Court for the Eastern District of Virginia, Microsoft and co-plaintiff Health-ISAC worked with Cloudflare, Coinbase, OpenAI, Railway, SpyCloud, the Shadowserver Foundation and TRM Labs to seize 50 websites and disable more than 150 additional domains, according to Microsoft. In the UK, the Metropolitan Police Service's cybercrime team arrested two men, aged 32 and 38, on 11 September 2026 on suspicion of operating the platform; both were released on police bail while the investigation continues, Microsoft said. The action is DCU's 40th court-authorised disruption and its first against what Microsoft called an "end-to-end AI-enabled cybercrime service."

Why it matters

The case is among the first documented instances of a commercial cybercrime service in which a chatbot ran the entire fraud pipeline — automated mailbox analysis, target selection, relationship mapping and impersonation-message drafting — rather than just producing convincing phishing text, and Microsoft's own investigators said they used reverse engineering and automated evidence-analysis tools to accelerate the counter-investigation. For organisations in the Caspian region and elsewhere that rely on Microsoft 365 identity systems, the device-code technique behind EvilTokens works regardless of geography, making Microsoft's advice to independently verify payment-change requests through a second channel relevant well beyond the countries named in its disclosure.