Researchers at the company looked into how malicious fine-tuning makes a model go rogue, and how to turn it back.

Duolingo learned the hard way that you can't joke your way through a backlash after its AI-focused announcement drew criticism.

Last January, Meta quietly deleted some earlier AI persona accounts after user backlash ... mishaps" are escalating into real harm. What's particularly striking is that no one asked for this ...

Character.AI has limitations. It can sometimes generate inaccurate information, called hallucinations, or responses that don't quite align with a character's intended persona. It's important to ...