Threat feed

TheMailer AI phishing kit

TheMailer: AI phishing kit rewrites emails until they bypass spam filters

Posted · Spectrum Delta Research Team

Summary

Varonis Threat Labs identified TheMailer, an unattributed self-hosted phishing kit sold on a cybercrime network since March 2026, which gives any operator with an API key and a target list machine-scale personalized lures and repeated spam-filter bypass (grade B/2). Its Inbox Loop sends a test message, checks the test inbox over IMAP, and if it landed in spam uses the connected model to rewrite the lure and retry until it reaches the victim inbox; every recipient gets a unique rewritten email, so content filters have no shared template to block. The kit writes lures from short prompts, clones real brand emails, tailors content to each mailbox provider, strips machine-writing patterns, hides links behind Google-domain redirects, and builds personalized PDF attachments with randomized metadata. A lifetime license sells for 3,000 to 6,000 dollars, and seller screenshots show a 24,829-email campaign and a 145,133-entry target list dominated by Japanese co.jp and ac.jp domains. Defenses should enforce SPF, DKIM and DMARC at p=reject, monitor lookalike and newly registered domains, and check full redirect chains at click time.

Breakdown

Countries

  • Japan
  • Germany
  • France
  • Russia

Targets

  • Industry: Japanese businesses and universities
  • Industry: Financial services

Technical exposure

  • Any organization with email: unique AI-written phishing lures at machine scale
  • Content filters degraded: no shared template or attachment hash to block
  • Links hidden behind Google and other trusted-domain redirects

Adversary

TheMailer seller (unattributed) · Varonis Threat Labs (grade B/2)

Procedures (3)

  1. AI Inbox Loop rewrites lures until spam-filter bypass

    Logs

    • Email gateway logs · repeated test sends from one sending infrastructure followed by a burst of unique personalized messages
    • Mail server logs · high-volume unique-body mail from recently warmed low-reputation sending accounts
  2. Per-recipient unique lures defeat template and hash blocking

    Logs

    • Email gateway logs · campaign bursts where every message body and every attachment hash differs
  3. Trust redirects hide links behind Google domains

    • T1566.002Phishing: Spearphishing Link
    • T1036.005Masquerading: Match Legitimate Resource Name or Location

    Logs

    • Web proxy logs · click-time redirect chains from mail links through google.com endpoints to credential pages
    • DNS logs · lookalike and newly registered domains resolving after trusted-domain redirects

Sources

Grades are source reliability (A–F) and information credibility (1–6).

Questions, detections or sightings of TheMailer AI phishing kit? Bring them to the community.