Leak Risks of File Uploads in xAI

WesEklund · x · 2026-07-14

This post highlights a security concern regarding Grok/xAI: when the model reads files for Q&A or code review, the file contents enter the LLM context in plaintext and are transmitted to the server.

The author points out that this means:

Related event: Grok Build/CLI accused of silently uploading entire codebases; xAI promises to delete data(47 posts)→

Original post →

More from Safety

Safety channel →