Chinese AI company Z.ai has disabled certain features of its ZCode coding assistant after a security issue involving the uploading of users' code repositories.

Reports said the problem was connected to a Codebase Indexing feature that was enabled by default. Users raised concerns that local repositories were being uploaded to overseas cloud infrastructure without sufficient consent. Z.ai subsequently apologized and said the issue had been patched.

Why is this significant?

AI coding assistants increasingly need access to source code.

That creates an obvious security challenge.

A developer may connect an AI tool to an entire repository containing:

Proprietary algorithms

API keys

Internal documentation

Customer information

Infrastructure configuration

Security credentials

If the tool uploads or processes that data unexpectedly, the organization can face serious exposure.

The problem isn't unique to Z.ai

The incident highlights a broader issue with AI developer tools.

Companies often focus on model quality and productivity while overlooking data flows.

Before deploying an AI coding assistant, organizations should know:

What data leaves the machine?

Where does it go?

How long is it retained?

Is it used for training?

Can administrators control the behavior?

What should enterprises do?

AI coding tools should be treated like any other third-party software handling sensitive data.

Security teams should evaluate:

Repository permissions

Network access

Data retention

Encryption

Telemetry

Training policies

Administrative controls

Developers should also avoid giving an AI assistant unrestricted access to production credentials or unrelated repositories.

The Z.ai incident is a useful reminder that AI security is not only about malicious models.

Sometimes the biggest risk is simply an AI tool having more access to company data than users realize.