Unauthorized AI Relay Service Compromises Developer Tools
An active relay service is bypassing security sandboxes in AI developer tools to intercept and reroute traffic for 12 organizations.
Updated on Oct. 29, 2026 in Artificial Intelligence
Live Poll
Is it worth using unofficial software tools if they expose your computer to potential security risks?
Since early 2026, an unauthorized relay service has compromised AI developer tools by replacing official credentials with relay-issued tokens. The service operates across four Asian markets, where it installs modified versions of OpenAI Codex and Anthropic Claude Code on organizational systems.
Why it matters
The service bypasses essential security controls to monitor or control developer interactions with AI models. This incident highlights the risks of executing unverified scripts in development environments, enabling attackers to intercept sensitive source code and proprietary logic.
The relay service replaces standard user API credentials with relay-issued tokens to intercept requests. It specifically disables command-line interface execution sandboxes while enabling network access for the write sandbox.
The players
Symantec
A cybersecurity firm that provides threat intelligence, enterprise security software, and endpoint protection.
OpenAI
An AI research lab and developer of the Codex model, which powers various coding assistant tools.
Anthropic
An AI safety and research company that produces the Claude series of large language models.
The details
Users are compromised via a one-line command that fetches a fileless script—a malicious payload that runs directly in system memory—using bash or PowerShell. The script logs users out of legitimate sessions, reconfigures CLIs to route traffic through relay infrastructure, and disables telemetry and error reporting features to mask its activity.
Timeline
The relay service began active operations in early 2026.
Symantec published a report on the relay service in October 2026.
The Tech Race
This relay service follows a pattern established by the 2020 SolarWinds supply chain attack by inserting unauthorized infrastructure into legitimate software workflows. It marks a shift toward targeting the developer ecosystem, where high-value proprietary source code is increasingly processed by AI tools.
Developers should immediately audit their CLI configurations and verify that their API credentials remain bound to official vendor services. The risk primarily affects organizations that pull and execute unverified scripts from non-official sources in their development environments.
The takeaway
Developers must treat CLI integration scripts as high-risk vectors for credential interception and sandbox bypass. Organizations should watch for future reports from security researchers tracking the evolution of AI-targeted supply chain attacks.
Further reading
For broader trends in AI security and development ecosystem risks, visit Artificial Intelligence.
Live Poll
Is it worth using unofficial software tools if they expose your computer to potential security risks?
