Microsoft Disrupts EvilTokens AI-Assisted Platform That Compromised 12,000 Accounts

Microsoft Leads Takedown of AI-Enabled Cybercrime Service

Microsoft announced it led a coordinated disruption of EvilTokens, an AI-assisted platform that compromised 12,000 Microsoft accounts across 10,000 organizations worldwide1

1

. The subscription-based scam platform charged an initial $1,500 fee and a recurring $500 monthly charge, offering cybercriminals a streamlined phishing-as-a-service solution3

3

. Through legal authorization from the U.S. District Court for the Eastern District of Virginia, Microsoft seized 50 websites and disabled more than 150 additional domains tied to the operation's infrastructure2

2

.

The Metropolitan Police Service arrested two men, aged 32 and 38, on September 11, 2026, on suspicion of offenses connected to the alleged operation of EvilTokens5

5

. Both suspects have been released on bail while the investigation continues, with their digital services and other items confiscated5

5

. The operation involved multiple partners including Health-ISAC, Cloudflare, Coinbase, OpenAI, Railway, SpyCloud, The Shadowserver Foundation, and TRM Labs3

3

.

How Device-Code Phishing Bypassed Traditional Security

EvilTokens exploited OAuth device code authentication, a legitimate process designed for input-constrained devices like TVs1

1

. The platform automated sending large volumes of spam containing malicious links that directed victims to webpages running hidden automation scripts1

1

. These scripts interacted in real time with Microsoft Entra, the identity provider, to generate device codes for enrolling attacker-controlled devices4

4

.

Victims saw device codes with instructions to enter them at Microsoft's legitimate device login portal, unknowingly granting attackers authenticated access to their Microsoft 365 applications2

2

. The platform's Node.js-based backend logic allowed operations to bypass traditional signature-based or pattern-based detection methods4

4

. This device-code phishing technique saw a 1,380% increase in attacks between July-December 2025 and January-April 2026, with over 50% linked to EvilTokens campaigns5

5

.

AI Chatbot Accelerated Business Email Compromise

At the center of EvilTokens was an AI chatbot that fundamentally changed the timeline of inbox compromise and payment fraud1

1

. The AI-boosted phishing tool could analyze 5,000 compromised emails at a time, identifying employees authorized to disburse large sums of money, their reporting managers, and convincing scenarios for transferring funds to attacker-controlled accounts1

1

.

The platform's AI capabilities extended beyond simple email analysis. It could identify trusted relationships, payment authorizations, sensitive responsibilities, and circumstances where fraud was most likely to succeed3

3

. The AI chatbot offered preset prompts to find wire-transfer discussions, locate vendor invoices, and determine the best people to impersonate3

3

. It could even draft messages impersonating trusted contacts to trick victims into taking action4

4

.

Microsoft tracking the threat actors as Storm-2992 found evidence that large portions of EvilTokens were "vibe coded," with AI helping creators build the platform itself5

5

. The platform drew capabilities from multiple AI models and ran like a well-organized startup with subscription pricing, customer support, management dashboards, and tools designed to move customers from account access toward financial exploitation5

5

.

Global Impact Across Industries and Regions

Victim organizations spanned wholesale distribution, construction, financial services, real estate, higher education, and healthcare sectors1

1

. The highest concentration of compromised accounts belonged to users in the United States, followed by Canada, the UK, Australia, India, and France1

1

. Microsoft observed 10 to 15 distinct campaigns launching every 24 hours since March 15, 20262

2

.

The platform was first documented in February 2026 and marketed through Telegram channels including @eviltokensadmin, @EvilTokens_bot, and @EvilTokensChannel3

3

. EvilTokens offered three products: B2B sender for $600, Office 365 capture link for $1,500, and SMTP sender for $1,0003

3

. The Office 365 capture link corresponded to the device-code phishing kit, granting affiliates lifetime access to the administration panel for viewing harvested Microsoft tokens3

3

.

What Organizations Should Watch For

Steven Masada, associate general counsel and general manager at Microsoft's Digital Crimes Unit, emphasized this marks DCU's first action against an end-to-end AI-enabled cybercrime service in nearly two decades of disruption operations2

2

. He warned that while EvilTokens infrastructure has been disrupted, the model it demonstrated will not disappear2

2

.

Microsoft advised organizations to assume that once an inbox is compromised, criminals may understand its contents in minutes, not days4

4

. Organizations should independently verify requests to change payment information, redirect funds, or approve unusual transactions through a trusted second channel1

1

. TRM Labs noted that EvilTokens packaged account takeover, AI-driven mailbox analysis, and fraud tooling into a single commercial service, lowering the expertise once needed to run business email compromise and invoice fraud at scale3

3

. Detective Inspector Serena D'Adamo stated the Met remains committed to holding people accountable who facilitate criminal enabling functions2

2

.