As cited
Copy frozen at (site build).
ai security
Shadowy AI Relay Disables OpenAI Codex, Claude Code Sandboxes
An intermediary service is distributing OpenAI's Codex and Anthropic's Claude with modified configurations to regions where these tools are not officially available, including mainland China. The relay disables sandbox protections and other security controls in these distributions. This circumvention allows access to artificial intelligence (AI) coding tools outside authorized markets while removing guardrails designed to prevent misuse.
Why it matters: Organizations and security teams in regulated industries should audit their AI tool usage to identify whether they are accessing these services through unauthorized channels, as disabled sandboxes create risk of code injection and data exfiltration.
- Source published
- First seen by Cybersecurity Tracker