When AI agents take unexpected actions

Learn how AI agents can take unexpected actions, understand the risks of goal misalignment, and apply practical safeguards for safer AI adoption.

Source: Australian Cyber Security Centre