Flitz.ai Flitz.ai

November 12, 2026

Mittwald Hit Again: Second AI-Assisted Attack Defaces Hoster's Own Website

A week after a breach through its AI-connected admin interface, German web host Mittwald was hit again overnight — this time attackers defaced its own corporate website. Here's what to check today. Ge...

Mittwald website defacement - second AI-assisted cyberattack

A week after a breach through its AI-connected admin interface, German web host Mittwald was hit again overnight — this time attackers defaced its own corporate website. Here's what to check today.

German web hosting provider Mittwald was struck by a second cyberattack overnight into 12 November 2026 — exactly one week after a first breach through its MCP server. This time the attackers went further: they took over and defaced Mittwald's own corporate website, and once again locked customers out of SSH access, the mStudio management platform and the customer center.

Mittwald told customers that hosted customer websites themselves remained online and reachable throughout the incident. But the company's internal infrastructure was hit hard enough that it described itself as being "in crisis mode" — and, notably, stated plainly that it is dealing with attackers who are clearly AI-assisted and professional. (heise.de)

What actually happened

The first attack, a week earlier, exploited a flaw in Mittwald's MCP server — the interface customers use to connect AI assistants directly to the mStudio hosting platform. That intrusion had already compromised 56 customer projects before it was contained. (mlq.ai)

According to Mittwald's own status page update on 12 November, the attackers resumed their campaign against the company's infrastructure and exploited a further vulnerability to compromise the corporate website itself. (borncity.com) In other words: this was not a one-off breach that got patched and closed. It was a return visit, by attackers who had clearly not been fully locked out the first time, going after a different part of the same organisation.

Why this matters beyond Mittwald

MCP servers — the connectors that let AI assistants read and act on admin interfaces, hosting panels, databases and code repositories — are being rolled out fast across the hosting and software industry. They are convenient. They are also, as this case shows twice over, a fresh and valuable target. An AI assistant with admin-level access is only as secure as the connector sitting between it and your infrastructure, and attackers are now professional enough to find and exploit those connectors repeatedly.

Swiss SMEs rarely run their own MCP servers directly, but many now use AI coding assistants, AI-connected support tools, or third-party platforms that quietly expose similar interfaces. If any system in your business lets an AI assistant log in, deploy code, or manage infrastructure, it deserves the same scrutiny as a VPN or an admin panel — because to an attacker, that's exactly what it is.

What to do today

  • If you or your developers use Mittwald's MCP server or mStudio: change your mStudio password immediately, enable two-factor authentication if you haven't already, and review all API keys and technical access credentials for signs of misuse. Mittwald has explicitly told affected customers to do this. (heise.de)
  • Inventory every AI-connected admin interface in your business. List any tool where an AI assistant, bot or automation script has standing credentials to a hosting panel, CMS, code repository or server. If you don't know the answer, that's the first problem to fix.
  • Turn on two-factor authentication everywhere an AI tool connects — not just for human logins. Treat AI assistant connectors as privileged accounts, not convenience features.
  • Rotate API keys on a schedule, not only after an incident. Keys that have been sitting unchanged for months are exactly what attackers look for.
  • Ask your hosting or IT provider directly whether they run any AI-to-admin connectors, and what happens if one is compromised at 2am on a Saturday.

The bigger picture

Mittwald being hit twice in a week, by attackers the company itself calls AI-assisted professionals, is a clear signal: AI-powered attacks on business IT are no longer rare events, they are daily business. Vulnerabilities in the tools connecting AI to admin systems are a growing, specific category of risk — one that didn't exist in this form two years ago and that most SMEs have not yet added to their security checklists.

This is exactly the kind of threat we'll keep tracking and reporting on, because the pace of disclosure is not slowing down. For Swiss SMEs who want help automating IT and business processes without creating this kind of exposure, TMC (The Mejlerö Company) in Engelberg supports businesses across Obwalden, Nidwalden, Lucerne and Zurich with practical IT support — from Microsoft 365 administration and endpoint protection to two-factor authentication setup — plus a free network analysis to see where your own exposure might be hiding.

Limited spots available this week

Get a FREE consultation

Get a FREE 1 hour consultation about your current application management setup, development process or any other tech setup problem.