xAI has clarified how Grok Build handles user data through three separate privacy tiers, with enterprise teams gaining access to zero data retention that processes requests without storing any trace data, code snippets, or session logs on xAI's servers. The announcement, posted by the official SpaceXAI account, addresses growing privacy concerns about AI coding tools and offers different protections depending on whether you're an enterprise team, API user, or individual subscriber.

Elon Musk directly addressed concerns about the zero data retention fine print on X, confirming that all user data uploaded to Grok prior to this clarification will be completely deleted. He acknowledged the policy language raised legitimate questions and framed the deletion as a precautionary measure, while noting that retaining some data is genuinely useful for debugging purposes.

What Are Grok Build's Three Privacy Tiers?

xAI has structured its privacy approach to serve different user needs and organizational requirements. The layered system reflects an attempt to balance security concerns with practical development needs across multiple customer segments.

• Zero Data Retention (Enterprise): An enterprise-level feature that processes API requests and responses entirely in real-time without writing any data to storage, meaning no trace data, code snippets, or session logs are persisted on xAI's servers. When active, API responses include an x-zero-data-retention header you can verify programmatically.
• API Key Coverage: All API key usage of Grok Build respects zero data retention when it is enabled for a team, meaning the same no-retention policy that covers interactive sessions also covers automated or programmatic requests made with API keys under that team account.
• CLI Privacy Mode: A narrower protection available to individual users through a /privacy command in the Grok Build command-line interface, which disables data retention at the session level even without enterprise zero data retention. This operates at the CLI layer rather than at the infrastructure level but gives individual users a meaningful opt-out without needing an enterprise contract.

How Can Individual Users Enable Privacy Protection?

For developers and teams without enterprise contracts, xAI has made privacy options available through its command-line tools. The /privacy command in Grok Build's CLI goes further than just data retention by isolating your session under a private GROK_HOME directory, disabling telemetry and feedback collection, and preventing codebase uploads to remote servers.

Individual users can also set a persistent veto on codebase uploads by adding disable_codebase_upload = true to their ~/.grok/config.toml file, which acts as a hard block on whole-repository uploads regardless of session settings. This gives developers granular control over what data leaves their local machines.

When Did Grok Build Launch and What Powers It?

Grok Build launched in early beta on May 25, 2026, available to SuperGrok and X Premium Plus subscribers. The tool runs on the grok-4.5 model under the hood, the same model accessible directly through the xAI API. After May 15, 2026, requests routed to the grok-code-fast-1 model were redirected to grok-build-0.1, consolidating the coding-focused model lineup.

The timing of these privacy clarifications comes as AI coding assistants face increasing scrutiny over data handling practices. Enterprise teams need to contact sales@x.ai to have zero data retention enabled for their organization, and once it is active at the team level, it applies across all members without requiring per-user toggles.

The distinction between enterprise-level infrastructure protection and CLI-level session protection matters for teams handling sensitive codebases. Whether the /privacy CLI command offers sufficient protection for non-enterprise teams working with proprietary code is a question worth asking your security team before committing production workloads to Grok Build.