Hallucinated domain in official OpenAI skill underscores AI supply chain risks
Summary
Researchers discovered a supply chain vulnerability in OpenAI's official skills repository where a skill intended to help deploy applications on Render inadvertently instructed production apps to send tokens to an unregistered domain, 'render-internal.com'. Upon registering the domain, the researchers found it was actively receiving tokens and API calls from real applications, highlighting potential exfiltration risks before OpenAI and Render resolved the issue.
IFF Assessment
The discovery of a supply chain vulnerability that could lead to the exfiltration of sensitive tokens represents bad news for defenders as it exposes a new attack vector.
Defender Context
This incident underscores the importance of thoroughly vetting and monitoring AI skills and plugins, treating them as critical components of the software supply chain. Defenders should be vigilant about potential vulnerabilities in AI-generated or AI-assisted code and dependencies, as even officially sanctioned tools can harbor risks.