A red-team experiment found an AI agent could autonomously exploit a vulnerability in McKinsey’s internal chatbot platform, exposing millions of conversations before the issue was patched.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results