xAI has open-sourced Grok Build after backlash over secretly uploading users’ repositories
xAI has released the full source code of Grok Build, its AI coding agent, after Cereblab researchers found that the tool was uploading users’ entire Git repositories to company controlled Google Cloud storage. The transferred data could include sensitive files such as SSH keys, password databases, private documents, and photos.
xAI disabled the file upload feature and said data storage had been turned off by default since July 12. Elon Musk also stated that previously uploaded user data would be permanently deleted and promised to open source the project after a security audit. The code is now available on GitHub under the Apache 2.0 license, although it was published as a single commit without previous Git history or pull requests.
The codebase contains approximately 844,530 lines of Rust covering Grok Build’s core agent workflow, tools, terminal interface, and plugin system. Code related to the disabled upload feature remains in the repository. Grok Build can read repository configurations, plan tasks, generate or modify code, execute shell commands, run in automated pipelines, and integrate with editors through the Agent Client Protocol.
- AI Chatbot
- Freemium
- Proprietary
Related news
External links
- Grok Build is Now Open SourcexAI News • Official source
- SpaceX open-sources Grok Build after data retention furoreThe Register
- xAI has open-sourced the Grok-Build source code following a massive data breachMezha
Comments
I think open-sourcing it was the right decision. Everyone makes mistakes, but being transparent and letting the community review the code is the best way to earn back trust. Hopefully this also pushes AI coding tools to be much clearer about what data leaves your computer.
They uploaded entire Git repositories, aka. Code projects including their change history etc. instead of only task required files. Quoting TheHackerNews: "With "Improve the model" turned off, Grok still uploaded the repository, and the server's own /v1/settings response kept returning trace_upload_enabled: true. That toggle governs whether your data trains the model. It does not govern whether your code leaves the machine."
Every big AI company constantly breaks copyright and laws and "accidentally" misleads. They seemingly "fixed" the issues, but re-giving them trust because of a PR-stunt, given (x)AI's reputation and again investing AI companies constantly seemingly justifying breaking ethics seems almost blatantly naive.