AI safety filters fail in long documents

Safety guardrails built to block harmful AI outputs quietly fail when text gets long, letting harmful content slip through undetected.

Also today

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