Microsoft dismantled EvilTokens, an AI-enabled cybercrime service tied to 12,000+ compromised e-mail inboxes
EvilTokens, a cybercrime service sold through Telegram, used an AI chatbot to analyze stolen inboxes. It identified payment staff and drafted impersonation messages for business e-mail compromise fraud. Microsoft says the service was tied to more than 12,000 compromised inboxes across over 10,000 organizations worldwide since launching in February. Microsoft and Health-ISAC filed a civil lawsuit in the Eastern District of Virginia, and the court authorized domain takedowns; more than 150 supporting domains were disabled. UK police arrested two men in connection with the alleged operation.
What the AI did
An AI chatbot built into the EvilTokens service analyzed stolen e-mail inboxes, picked out the staff who handle payments, and wrote messages pretending to be trusted contacts to trick them into sending money. The service reportedly misused AI services from Groq and OpenAI.
First reported September 28, 2026 · Added to the register September 29, 2026 · 1 source
- Compromised inboxes
- More than 12,000
- Organizations affected
- Over 10,000 worldwide
- Developer
- Groq and OpenAI services
- Deployer
- EvilTokens
- Affected
- Financial services, healthcare, construction and others
- When it happened
- February 2026
- First reported
- September 28, 2026
What this means for you
Could this affect you?
If you use Microsoft 365 e-mail. Organisations using Microsoft 365 e-mail were exposed, with more than 12,000 inboxes compromised across over 10,000 organisations in sectors including financial services, healthcare, construction, real estate, wholesale distribution and higher education, and payment staff were targeted for impersonation.
What to check
- Train staff to reject unexpected device-code sign-in prompts.
- Consider restricting the device-code sign-in method in your e-mail system.
- Revoke access tokens and end active sessions after a suspected compromise, rather than relying on password resets alone.
- Confirm any request to change payment details through a separate, known contact channel.
Areas of your AI programme this touches
Timeline
- February 2026Happened
- September 28, 2026First reported
Every fact and its source (6)
- Service launchFebruary“Since launching in February”[1]
- Arrests in UKTwo men arrested on Sept. 11“arrested two men on Sept. 11”[1]
- Domains disabledMore than 150“disabled more than 150”[1]
- Subscription price$500 per month“$500 monthly subscription”[1]
- Lawsuit filedCivil suit by Microsoft and Health-ISAC“Microsoft and Health-ISAC filed a civil lawsuit in the U.S.”[1]
- Court venueEastern District of Virginia“Eastern District of Virginia”[1]
Sources
- AI Platform Behind 12 , 000 E - mail Breaches Dismantled -- THE Journalthejournal.com · September 28, 2026
How this record is classified. Severity C (5/5): real harm at scale, or an autonomous system acting against people or institutions. OECD level: serious incident, an event in which an AI system led to serious harm. Evidence: Confirmed, meaning a company, official or court statement.
The facts in At a glance and in Every fact and its source are quoted from the sources listed. The summary, What the AI did and What this means for you are written from those sources and checked against them automatically before publication. Records reflect what has been disclosed, not everything that has happened. Incidents are found through news feeds and the GDELT Project.
