Incident · Nov 2026 · Excess permissions

Mittwald MCP server

A vulnerability in the company’s AI-assistant-related Model Context Protocol server let an attacker move through the platform as a normal user before abusing the flaw. The incident reportedly affected 56 customer projects.

What happened

A vulnerability in the company’s AI-assistant-related Model Context Protocol server let an attacker move through the platform as a normal user before abusing the flaw. The incident reportedly affected 56 customer projects.

The attacker first tested the weakness, then used it to operate inside the platform with ordinary-user access instead of the restricted role the service was meant to enforce.

Because the assistant server could reach more data and actions than it should have, the flaw opened a path to customer projects that were not meant to be exposed.

The incident was not just a theoretical bypass: the reporting says real customer projects were impacted.

Mittwald’s response, as summarized in the incident reporting, was to acknowledge the abuse of the flaw after the compromise was traced.

Source: Oh My AI ↗. Follow it for the full account.

The cause: excess permissions

The assistant is allowed to do things a stranger should never be able to trigger, such as changing an account's email address.

Framework entries that cover it: OWASP LLM06: Excessive Agency

The solution
  1. 1

    What would have caught this one

    A strict privilege audit on the assistant server, with role-bound access checks on every action, would have blocked the abuse.

  2. 2

    Prevent the whole class

    List every action the assistant can take and remove what it does not need. Put identity checks in front of anything sensitive.

  3. 3

    Test for it

    Try to trigger each action as an unverified stranger.