News Artificial Intelligence

Innocent-looking AI reasoning can make bad behavior harder to catch

One AI can monitor another, but the safeguard may miss signs of trouble

AI agents can check each other's reasoning to look for bad behavior

One way to keep tabs on AI agents is to have another AI monitor their written reasoning for signs of trouble. New research suggests that safeguard can falter when the reasoning is the main clue that something is wrong.

J Studios/DigitalVision/Getty Images

By Sascha Brodsky

Safety concerns about AI are growing, and now, a common method for catching misbehaving systems could have flaws, new research suggests.