Cyber Risks · The Hacker News

Docker Fixes Critical Ask Gordon AI Flaw Allowing Code Execution via Image Metadata

Summary: Researchers at Noma Labs disclosed a now-patched critical flaw, codenamed DockerDash, in Docker's Ask Gordon AI assistant that let a single malicious metadata label in a Docker image trigger a three-stage attack: Gordon reads the instruction, forwards it to the MCP Gateway, which executes it through MCP tools with zero validation. Docker addressed it in version 4.50.0 in November 2025, which also resolved a separate prompt injection flaw found by Pillar Security involving tampered Docker Hub repository metadata.

Why it matters: The exploit needs no vulnerability in your code, only a LABEL field in an image you pull, and the same flaw yields remote code execution on cloud and CLI systems or exfiltration of installed tools, container details, Docker configuration, mounted directories, and network topology on Docker Desktop.

Read the original at The Hacker News ↗

Topics
AI Security
Related