Threat Intelligence

Unofficial AI Relay Offers Coding Tools in Restricted Regions, With Security Protections Switched Off

Dark Reading · 29 Oct 2026

Key Takeaway Only use AI coding tools through the vendor's official channels and approved accounts, so built-in safeguards such as sandboxes stay in place.

A report from Dark Reading describes an unusual service that acts as a relay for popular AI coding tools. It gives users in mainland China and other regions access to OpenAI Codex and Claude Code, even though OpenAI and Anthropic do not officially sell their products there.

The concern is what has been removed along the way. According to the report, the service operates with key security features turned off, including the sandboxes designed to limit what these coding tools can do on a user's system. The operator of the service remains unclear, and the report describes it as enigmatic.

The summary available does not give further technical detail, so the full extent of the risk is not yet known. Even so, the case is a useful reminder for businesses. AI coding assistants can read, write and run code, so the safeguards around them matter. Using unofficial or third-party routes to reach these tools can strip away those protections and add parties you cannot vet.

AI security OpenAI Codex Claude Code supply chain risk

Summarised by CISO AI from Dark Reading, written with Claude Sonnet 5.5. We link back to every original so you can read it yourself.