Musk Pledges to Delete Grok Build User Data as Tool Uploaded Entire Codebases
Updated
Updated · The Verge · Jul 14
Musk Pledges to Delete Grok Build User Data as Tool Uploaded Entire Codebases
3 articles · Updated · The Verge · Jul 14
Summary
Elon Musk said all data previously uploaded by Grok Build will be “completely and utterly deleted” after researchers found the coding tool had sent users’ repositories to Google Cloud.
Cereblab reported the CLI was packaging entire codebases—including excluded files and secrets deleted from history—far beyond the retention seen in rivals such as Claude Code.
Tests on Monday showed SpaceXAI’s servers returning a “disable_codebase_upload: true” flag, and the upload behavior no longer triggered after the company shut the feature off.
SpaceXAI had pointed users to a /privacy command, but Cereblab said that setting only toggled per-session retention and was not the control that stopped the uploads.
Dr. Lukasz Olejnik called the retention excessive, warning exposed data could include proprietary source code, security flaws, personal data, infrastructure details and credentials.
Will the Grok Build scandal force enterprises to abandon cloud AI for more secure, on-premise alternatives?
After this data leak, can developers ever trust AI coding assistants with their source code again?
Is the massive data collection by AI tools a bug, or an unavoidable feature of creating powerful AI?
July 2026 Grok Build Data Leak: xAI’s Secret Code Repository Uploads and the Future of AI Privacy
Overview
In July 2026, a security researcher publicly revealed that xAI's Grok Build coding tool was uploading users' entire code repositories to xAI's cloud without clear user consent. This undisclosed data transfer sparked widespread concerns about privacy and the security of intellectual property, as sensitive materials like source code, project files, and confidential business logic could have been exposed. The controversy quickly grew, highlighting serious flaws in Grok Build's privacy controls and raising urgent questions about user trust, transparency, and the need for stronger data protection in AI development.