Sandbox Flipped to 'Danger-Full-Access': Unauthorized AI Relays Create Risk for Organizations

An unauthorized relay service has been installing legitimate OpenAI Codex and Anthropic Claude Code CLI tools on enterprise systems, reconfiguring them to route requests through third-party infrastructure. This activity, active since early 2026, primarily affects organizations in mainland China, South Korea, Taiwan, and Malaysia across multiple industrial sectors. The relay disables execution sandboxes, enables network access, replaces API credentials with relay-issued tokens, and disables telemetry, granting operators full user privilege command execution and access to sensitive data. No abuse evidence is currently reported.

• Variola: From ClickFix to a Modular ImplantMalware2026-10-16
• Kimsuky attacks targeting Ukrainian experts in diplomacy, security, and international policyCampaign2026-10-13
• August 2026 Threat Trend Report on APT Attacks (South Korea)Malware2026-10-02
• Star Blizzard refines phishing and malware delivery with the RedFlick techniqueMalware2026-09-30
• Attackers Abuse ChatGPT Custom GPTs to Deliver RAT via ClickFixMalware2026-09-29