India, Sept. 9 -- As enterprises connect AI assistants to email, cloud storage, collaboration platforms, source-code repositories and internal applications, security can no longer stop at the model or user interface. The Check Point Research finding shows how a shared internal service used for package delivery could effectively become a covert communication channel between otherwise isolated ChatGPT execution environments.

The rapid adoption of AI assistants inside enterprises is creating a new infrastructure security challenge, how to isolate AI workloads when the systems supporting them still need to share common services.

Check Point Research has disclosed a vulnerability in ChatGPT's code-execution environment that, according to the...