TheMailer – Self-hosted Phishing Platform that Runs on Claude and ChatGPT

Blog WriterCybersecurity News - Original News Source is cybersecuritynews.com


TheMailer, a self-hosted phishing platform identified by Varonis Threat Labs, brings commercial AI models into almost every stage of an email attack.

Rather than only fixing grammar or writing convincing messages, it uses AI to build lures, change templates, test delivery, and manage replies. Operators supply their own API keys to connect services including Claude, ChatGPT, Grok, Gemini, Mistral, and DeepSeek.

According to the research, TheMailer has been advertised on a cybercrime network since March 2026. Its sales thread has attracted more than 24,000 views, although that figure does not confirm customer numbers or successful attacks. Advertised lifetime licenses range from $3,000 to $6,000, depending on the package and whether it includes AI modules.

TheMailer Forum Sales Advertisement (Image Source: Varonis)

Buyers receive installation help and support, while supplying their own mail servers, proxies, and target lists. A Telegram sales channel and automated cryptocurrency payment bot help handle purchases.

The setup reduces the technical work needed to start a campaign by packaging email delivery and AI tools in an interface that resembles ordinary marketing software.

The approach builds on tools such as SpamGPT, an AI-powered phishing platform, which combines message generation, mail account management, delivery testing, and campaign tracking.

TheMailer takes that model further by linking AI rewriting directly to feedback about where test messages land. That connection turns content creation into an ongoing process rather than a single preparation step.

Its most notable feature, Inbox Loop, sends a test email and checks a connected mailbox through IMAP to determine whether the message reached the inbox or spam folder.

When a test lands in spam, the platform asks AI to rewrite the template and tries again. This tests delivery conditions, but does not guarantee access to every recipient’s inbox.

TheMailer Campaign Management Dashboard
TheMailer Campaign Management Dashboard (Image Source: Varonis)

The Varonis Threat Labs report describes an AI Letter Builder that creates HTML email from short prompts. It can also import genuine messages from a connected mailbox and use them as templates.

This gives attackers an easy starting point for copying familiar brand layouts and language without designing each lure themselves.

Templates and subject lines can be rewritten for individual recipients. One example uses a Robinhood account warning with a changing reference number. Other presets turn messages into security alerts or routine transaction notices.

Options advertised as removing “AI patterns” aim to make the wording seem human, but the report does not establish that these options reliably defeat detection.

An AI Autopilot module groups recipients by mailbox provider and applies provider-specific sending limits and content guidance. Supported groups include Gmail, Microsoft, GMX, Yandex, and French providers.

A screenshot shows a list containing 145,133 entries, largely Japanese business and university domains. That is evidence of a displayed target list, not proof that those organizations were compromised.

List containing 145,133 entries (Image Source: Varonis)

The platform also advertises account warm-up, delivery testing, and reply management. AI WarmUP appears designed to exchange messages between sending accounts so they look established.

Inbox Finder tests delivery settings against a seed mailbox. A separate reply manager gathers responses and offers AI sorting and answering, extending automation beyond the initial message.

TheMailer’s attachment tools convert HTML templates into personalized PDFs, including filenames based on recipient names. Embedded links and changing metadata produce files with different hashes, making simple file blocklists less useful.

Other features create SVG redirects and calendar invitations. Branding shown in sample lures includes Robinhood and IONOS, alongside financial alerts and hotel booking callback themes.

The delivery features include randomized headers, sender profiles, and redirects that place trusted Google domains ahead of another destination.

These features can complicate review, but a familiar opening domain does not make the final page safe. Likewise, changing a message or file hash does not remove the phishing request or stop deeper analysis of its links and behavior.

The main defensive lesson is to assess intent, identity, and behavior rather than writing style alone. As explained in Cybersecurity News’ coverage of AI-written phishing detection, identifying machine-written text is different from identifying a malicious request.

Teams should check sender history, authentication results, reply-address mismatches, redirect chains, and attachment behavior.

Unexpected requests for passwords, payments, or verification codes deserve review even when every sentence looks natural. Staff should verify sensitive requests through a known contact channel rather than details supplied in the message.

What Are Leading Organizations Doing About AI and Cyber Risks? Find Out. View Agenda!

The post TheMailer – Self-hosted Phishing Platform that Runs on Claude and ChatGPT appeared first on Cyber Security News.