Experts find flaws in tests that check AI safety and effectivenesstheguardian.com3 points·Cynddl··0 commentsOpen articleSaveView on HN