A German web hosting company revealed over the weekend that its systems had been infiltrated by attackers who allegedly leveraged artificial intelligence during their breach. The intrusion targeted a feature built to let customers integrate AI assistants into their accounts, according to the firm's public disclosure.
Mittwald CM Service, which operates out of North Rhine-Westphalia and serves web agencies, announced Saturday that 56 customer projects fell victim to the attack after the intruders remained undetected for eight days. The breach exploited a vulnerability in the company's Model Context Protocol (MCP) server, which powers mStudio, Mittwald's platform for website and hosting management.
Managing director Robert Meyer attributed the breach to a flaw in either Mittwald's implementation or the underlying open-source MCP libraries. Meyer noted that the attackers appeared to deploy AI during their intrusion, pointing to system logs that demonstrated unusually rapid movement through Mittwald's infrastructure. "The logs show that he worked his way through our system extremely quickly," Meyer stated, adding that "in some cases, only seconds passed between steps that would take significantly longer for a human attacker."
Mittwald offered no additional corroboration of AI involvement in the attack. When pressed to clarify whether the company could distinguish between detected AI tools and conventional automation scripts—both of which enable attackers to operate faster than manual intervention—the firm did not provide immediate comment.
According to Mittwald's account of events, the attacker first tested the MCP vulnerability in October, amid widespread efforts to patch the PHP2Shell flaw. Meyer noted that the intruder "disguised his access attempts and carried out various diversionary tactics."
External forensic specialists have been engaged to support the investigation. "We have informed the relevant authorities," Meyer said. Following forensic review of the 56 affected projects and multiple scans of Mittwald's own infrastructure, the company reported as of Friday that no further attacker activity could be detected.
Meyer signalled that the company intends to release a comprehensive account of the incident once analysis and recovery efforts conclude. "First the PHP2Shell vulnerability, now the attack on our systems. AI attacks will become more frequent in the future," Meyer wrote. "We want to be transparent about this, and therefore we will publish a detailed account of the events next week, as soon as our analysis and recovery are complete. We will be working on this throughout the weekend and beyond."



