The Hallucination HQ

A free daily newsletter that makes AI news actually fun to read. Five stories, zero jargon.

Get it in your inbox every morning

Here's what it looks like

Google Teaches AI to Forget Things on Purpose

Google DeepMind published a paper on "machine unlearning" — basically teaching AI models to selectively forget specific training data without retraining from scratch. Think of it as Ctrl+Z for a neural network's brain: if that data shouldn't be there (or breaks privacy law), just undo it. The technique could help companies respond to GDPR deletion requests without the nuclear option of rebuilding entire models.