# TheMailer – Self-Hosted AI Phishing Platform Using Claude, ChatGPT and Other LLMs

> Varonis Threat Labs reports TheMailer, a self-hosted phishing platform sold on a cybercrime network since March 2026 that drives commercial LLMs (operator-supplied API keys for Claude, ChatGPT, Grok, Gemini, Mistral, DeepSeek, OpenRouter) through lure creation, template rewriting, inbox-placement testing and reply handling. It is sold as a lifetime license for roughly $3,000-$6,000.

- **Published:** 2026-11-17T17:25:28Z
- **Last reviewed:** 2026-11-17T17:25:28Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-6262
- **Card image:** https://intel.threadlinqs.com/og/threat/TL-2026-6262.png
- **ID:** TL-2026-6262
- **Severity:** MEDIUM
- **Category:** PHISHING
- **Status:** ACTIVE
- **Detections:** 9 (1 shown free below) · **IOCs:** 15 (2 shown free below)

## Description

TheMailer is a self-hosted, AI-enabled phishing and spam platform advertised on a cybercrime network since March 2026. Sales are run through a Telegram channel and an automated cryptocurrency payment bot that delivers licenses without seller involvement, with free installation and support. The sales thread has more than 24,000 views, which does not confirm customers or successful attacks. Lifetime licenses are advertised at $3,000 to $6,000 depending on the deal and whether AI modules are included. Buyers supply their own SMTP servers, proxies and target lists, and bring their own API keys for Claude, ChatGPT, Grok, Gemini, Mistral, DeepSeek and OpenRouter. The configuration page reportedly advises routing AI calls through SOCKS5 or residential proxies to avoid provider blocks on cloud IP ranges.

The platform is organized as modules. Inbox Loop sends test emails, checks via IMAP whether they landed in the inbox or spam folder, and has an LLM rewrite and retest until the message is delivered. AI Letter Builder generates HTML emails from text prompts and can import real brand emails from a connected mailbox as templates; AI Template Generator offers presets such as 'security or account alert' and 'legitimate transactional' and options to replace spam triggers and remove AI-like patterns, and personalizes templates and subject lines per recipient. AI Autopilot groups recipients by mailbox provider (Gmail, Microsoft, GMX, French ISPs, Yandex) and tailors content using per-provider spam-filter notes and sending limits. AI WarmUP runs automated send-and-reply exchanges between new sending accounts to build reputation. Inbox Finder tests combinations of SMTP accounts, headers, encoding, DKIM settings and content against a seed mailbox. Inbox Reply Manager collects target replies and uses AI to sort and answer them. File Templates builds personalized PDF attachments (filenames such as Invoice_%name%.pdf), SVG redirects and calendar invites; embedded metadata and links vary so no two generated files share a hash.

Delivery features include direct-to-MX sending over SOCKS5 proxies, randomized headers, content and sender profiles, deployment of proxies on AWS, 'trust redirects' that place Google domains ahead of the final destination, and open-redirect parsing and newer default AI models added in recent changelogs. Sample lures shown by the seller impersonate Robinhood ('Unusual activity' subject with randomized ID and an SVG lure reading 'Please call to customer support') and IONOS ('Kundenservice IONOS'), plus financial alerts, account security notices and hotel-booking callbacks. A seller screenshot shows a 24,829-email campaign and an MX check of a 145,133-entry target list, almost entirely Japanese .co.jp and .ac.jp business and university domains; this is a displayed target list, not evidence of compromise.

Varonis previously reported on SpamGPT, an AI-powered spam-as-a-service platform sold on underground forums (KaliGPT assistant, SMTP/IMAP checker, inbox check module, SMTP rotation, abuse of AWS and SendGrid), and says TheMailer 'goes further' by building AI into the whole platform; the sources do not document a direct code or actor link. The sources provide no CVEs, IOCs such as IPs/domains/hashes, or actor attribution. Recommended defenses: DMARC/SPF/DKIM at p=reject, monitoring lookalike and newly registered domains, click-time analysis of the full redirect chain, and intent/behavior-based analysis (sender history, authentication results, reply-address mismatches, redirect chains, attachment behavior) rather than content or hash matching. Severity is an analyst judgment (no CVSS, no confirmed exploitation).

## MITRE ATT&CK

- T1588.007 Obtain Capabilities: Artificial Intelligence
- T1585.002 Establish Accounts: Email Accounts
- T1583.003 Acquire Infrastructure: Virtual Private Server
- T1566.001 Phishing: Spearphishing Attachment
- T1566.002 Phishing: Spearphishing Link
- T1566.004 Phishing: Spearphishing Voice
- T1204.001 User Execution: Malicious Link
- T1204.002 User Execution: Malicious File
- T1684.001 Impersonation
- T1036 Masquerading
- T1027 Obfuscated Files or Information
- T1684.002 Email Spoofing
- T1090 Proxy

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**TheMailer Brand-Impersonation Lure (IONOS / Robinhood) from Non-Brand Sender** (`TL-2026-6262-SPL-001`)

Detects inbound mail carrying the TheMailer sample-lure indicators published by Varonis (subject 'Unusual activity' impersonating Robinhood, 'Kundenservice IONOS' impersonating IONOS support, 'Please call to customer support' callback text) sent from a domain that does not belong to the impersonated brand, fanned out to 3 or more recipients. Covers T1656 impersonation, T1672 email spoofing and T1566.004 callback phishing delivered by the TheMailer AI phishing platform (TL-2026-6262).

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1656, T1672, T1566.004
- **Data sources:** o365_message_trace, defender_emailevents
- **Severity:** medium
- **Confidence:** medium

### Splunk SPL

```spl
index=email sourcetype="ms:o365:reporting:messagetrace" (Subject="*Unusual activity*" OR Subject="*Kundenservice IONOS*" OR Subject="*Please call to customer support*") | eval sender_domain=lower(mvindex(split(SenderAddress,"@"),1)) | eval impersonated_brand=case(like(lower(Subject),"%unusual activity%"),"Robinhood",like(lower(Subject),"%ionos%"),"IONOS",true(),"CallbackLure") | where NOT (impersonated_brand="Robinhood" AND match(sender_domain,"(^|[.])robinhood[.]com$")) AND NOT (impersonated_brand="IONOS" AND match(sender_domain,"(^|[.])ionos[.](com|de|fr|es|it|co[.]uk)$")) | stats count AS messages dc(RecipientAddress) AS recipients values(Subject) AS subjects values(FromIP) AS from_ips min(_time) AS first_seen max(_time) AS last_seen BY SenderAddress sender_domain impersonated_brand | where recipients>=3 | convert ctime(first_seen) ctime(last_seen)
```

### Microsoft KQL

```kql
EmailEvents
| where Timestamp > ago(24h)
| where EmailDirection == 'Inbound'
| where SenderDisplayName has 'Kundenservice IONOS' or Subject has 'Unusual activity' or Subject has 'Please call to customer support'
| extend ImpersonatedBrand = case(SenderDisplayName has 'IONOS' or Subject has 'IONOS', 'IONOS', Subject has 'Unusual activity', 'Robinhood', 'CallbackLure')
| where not (ImpersonatedBrand == 'IONOS' and SenderFromDomain matches regex @'(^|[.])ionos[.](com|de|fr|es|it|co[.]uk)$')
| where not (ImpersonatedBrand == 'Robinhood' and (SenderFromDomain =~ 'robinhood.com' or SenderFromDomain endswith '.robinhood.com'))
| summarize Messages = count(), Recipients = dcount(RecipientEmailAddress), Subjects = make_set(Subject, 5), SenderIPs = make_set(SenderIPv4, 5), FirstSeen = min(Timestamp), LastSeen = max(Timestamp) by SenderFromAddress, SenderDisplayName, SenderFromDomain, ImpersonatedBrand
| where Recipients >= 3
```

### Sigma

```yaml
title: TheMailer Brand Impersonation Lure - IONOS and Robinhood Phishing
id: 7c1e4b3a-5d92-4f08-a6b1-2e9d0c3f8a11
status: experimental
description: Detects inbound email using the TheMailer sample lure indicators (Kundenservice IONOS display name, Unusual activity subject, Please call to customer support callback text) from senders outside the impersonated brand domains. Field names follow Microsoft Defender EmailEvents.
references:
    - https://www.varonis.com/blog/themailer-ai-phishing-kit
author: Threadlinqs Team
date: 2026/11/17
tags:
    - attack.initial-access
    - attack.t1566.004
    - attack.defense-evasion
    - attack.t1656
    - attack.t1672
logsource:
    category: email
    product: m365
detection:
    selection_ionos:
        SenderDisplayName|contains: 'Kundenservice IONOS'
    selection_robinhood:
        Subject|contains: 'Unusual activity'
    selection_callback:
        Subject|contains: 'Please call to customer support'
    filter_legit_senders:
        SenderFromDomain:
            - 'ionos.com'
            - 'ionos.de'
            - 'robinhood.com'
    condition: 1 of selection_* and not 1 of filter_*
falsepositives:
    - Legitimate Robinhood or IONOS notifications relayed through a third-party mail service provider
    - Other vendors using the generic subject Unusual activity
level: medium
```

**Known false positives:**

- Legitimate Robinhood or IONOS notifications relayed through a third-party transactional mail provider on a non-brand domain
- Other vendors that use the generic subject 'Unusual activity' for genuine sign-in alerts

**Tuning notes:** Add your organisation's approved brand-relay domains to the exclusion filters, raise the recipients threshold if generic 'Unusual activity' subjects are noisy, and extend the subject/display-name list with brands your users actually transact with. Sample lures are seller-provided examples; TheMailer generates per-recipient subjects, so treat this as a seed list and pair it with sender-reputation analysis.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- TheMailer-Style Bulk Lure Campaign with Failed Authentication and Sender Domain Mismatch (locked: unlock with TL Creds)
- TheMailer Google Trust-Redirect Link Clicked from Email (locked: unlock with TL Creds)
- Commercial LLM API Calls from Non-Browser Process (TheMailer AI Back-End Pattern) (locked: unlock with TL Creds)
- TheMailer SVG, Calendar-Invite or Personalised Invoice Attachment Delivered by Email (locked: unlock with TL Creds)
- Outlook-Opened SVG Attachment Launches Browser (TheMailer SVG Redirect Lure) (locked: unlock with TL Creds)
- SOCKS5 Proxy Egress Consistent with TheMailer Proxy-Routed Sending and AI Calls (locked: unlock with TL Creds)
- TheMailer Personalised Invoice PDF Opened Then Link Launched in Browser (locked: unlock with TL Creds)
- TheMailer Direct-to-MX SMTP Fan-Out from a Single Internal Host (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sample indicators (2 of 15)

Defanged, TLP:CLEAR.

- Domain: `co[.]jp`
- Domain: `ac[.]jp`

Full set of 15 indicators unlocks with TL Creds: https://threadlinqs.com/tl-creds

## Sources

- [Meet TheMailer: An AI Phishing Kit (Varonis Threat Labs)](https://www.varonis.com/blog/themailer-ai-phishing-kit)
- [TheMailer – Self-hosted Phishing Platform that Runs on Claude and ChatGPT (Cyber Security News)](https://cybersecuritynews.com/themailer-ai-phishing-platform/)
- [SpamGPT: The AI Tool Elevating Email Security Threats for Enterprises (Varonis)](https://www.varonis.com/blog/spamgpt)
- [SpamGPT phishing tool (Cyber Security News)](https://cybersecuritynews.com/spamgpt-phishing-tool/)
- [New AI-Driven Phishing Platform Automates Attack Campaigns (KnowBe4)](https://blog.knowbe4.com/new-ai-driven-phishing-platform-automates-attack-campaigns)
- [SpamGPT Amps Up Enterprise Email Security Threats (Security Boulevard)](https://securityboulevard.com/2025/09/spamgpt-amps-up-enterprise-email-security-threats/)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-6262
