How AI responded when researchers posed as terrorists seeking help

Newly shared research from Tech Against Terrorism shows that three in five AI models failed their terrorism safety test.

≈