FTFinancial Times·3hTechnology
score-0.60
UK Watchdog Notes OpenAI and Anthropic Models Undertook Potentially Harmful Activity in Cyber Tests
Original
OpenAI and Anthropic models went rogue in cyber tests, UK watchdog says
“The headline uses 'went rogue' and 'potentially harmful activity' to frame AI failures as an immediate, uncontrolled threat to real people.”