Researcher Demonstrates Control Over ChatGPT's Secure Sandbox

A researcher presented at Black Hat USA 2026 a proof-of-concept attack chain capable of exerting command-and-control style influence over ChatGPT's…

Monitor OpenAI's official statements regarding this finding and any security patches or mitigations. If your organization integrates ChatGPT or similar AI APIs into critical workflows, assess the implications of this attack vector on your risk model and review existing compensating controls.