A security researcher found 35,000 apps built with xAI's Grok Build agent left publicly exposed, including data belonging to SpaceX.
A researcher discovered that roughly 35,000 applications generated by Grok Build, xAI's AI coding agent, were left wide open and accessible to anyone who looked. The exposed data reportedly includes confidential information tied to SpaceX.
The scale of the exposure points to default configurations in agentic coding tools that fail to enforce basic access controls on generated apps.
As coding agents move from demos into production use by serious enterprises, insecure defaults become a direct liability, not a hypothetical one. Any team shipping AI-generated infrastructure needs an audit step before deployment, and vendors need to own secure-by-default design.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →