AI ethics labels hide misaligned moral reasoning

AI models often agree with human ethical verdicts but for different moral reasons, making label-based alignment tests misleading.

Also today

Today's top five AI stories, one line each, at 7am Eastern. Free. One-click unsubscribe.