Two major AI coding assistants have been caught uploading developers' entire code repositories to external cloud servers without explicit user consent, exposing a critical privacy gap in tools that enterprises are rapidly adopting. The incidents reveal how quickly AI development tools can be deployed without adequate safeguards for sensitive intellectual property.

What Privacy Incidents Have Occurred With AI Coding Tools?

Chinese artificial intelligence company Z.ai disabled several features of its ZCode coding assistant after a default setting was discovered sending users' local code repositories to Alibaba Cloud servers in China without their consent. The company subsequently disabled the workflow responsible for generating and uploading local repository snapshots and opened its codebase for public scrutiny. Months earlier, SpaceXAI's Grok Build coding CLI was found uploading entire Git repositories to a Google Cloud Storage bucket under its control.

These parallel discoveries highlight a troubling pattern. Developers using these tools may not have fully understood that their proprietary code was being transmitted to external cloud infrastructure. Git repositories contain far more than visible code: they include commit histories, configuration files, API keys, internal documentation, and other sensitive information that enterprises treat as confidential.

Why Should Enterprises Be Concerned About Code Privacy?

The incidents raise fresh concerns for enterprises over how AI tools handle sensitive source code. For large organizations, source code represents years of investment, competitive advantage, and intellectual property. When an AI tool uploads this code to cloud services without explicit consent, multiple risks emerge simultaneously.

First, there is the risk of unauthorized access if cloud storage buckets are misconfigured or exposed. Second, code may persist in cloud storage longer than users expect, creating a prolonged window of exposure. Third, there is geopolitical risk, particularly when code is uploaded to servers in countries with different legal frameworks and government data access policies. Additionally, many organizations operate under regulatory requirements like GDPR or HIPAA that restrict where sensitive data can be stored and processed. Uploading code without explicit consent may violate these compliance obligations.

How to Protect Your Code When Using AI Coding Tools

• Review Default Settings: Before using any AI coding assistant, carefully examine default configuration options, particularly those related to data transmission, cloud uploads, and repository access. Many tools enable data sharing by default, requiring users to actively opt out rather than opt in.
• Monitor Network Activity: Track outbound network connections from your development environment to identify unexpected data transfers to external cloud services. Network packet analyzers can reveal whether code is being uploaded without your knowledge.
• Apply Principle of Least Privilege: Grant AI coding tools only the minimum permissions necessary to function. Avoid giving them blanket access to entire repositories or directories containing credentials, configuration files, or proprietary algorithms.
• Use Isolated Environments for Sensitive Work: For projects involving highly confidential code, consider using AI coding assistants in isolated development environments that cannot access your main repository or connect to external cloud services.
• Establish Enterprise Approval Processes: Organizations should create clear policies governing which AI coding tools developers can use, what types of code can be processed, and what approval processes must occur before adoption.

The broader lesson is that convenience and innovation must not come at the cost of security and privacy. As AI tools become more integrated into the development workflow, the industry needs stronger standards for transparent data handling, explicit user consent, and privacy-by-design principles. Until those standards are established and enforced, developers and enterprises must remain vigilant about what data they share with AI coding assistants and where that data ultimately ends up.

Z.ai's decision to disable the problematic workflow and open its codebase for public scrutiny represents a step toward accountability, but it came only after the privacy breach was discovered. Developers should treat AI coding tools with the same security scrutiny applied to other third-party software, reading privacy policies carefully, testing default configurations in non-production environments, and maintaining visibility into data flows.