Experts find flaws in tests that check AI safety and effectiveness theguardian.com 3 points by Cynddl 7 hours ago