Safety gates for AI that hold under any generator

A new theory proves when AI safety filters stay reliable regardless of what model or policy is generating actions.

Also today

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