▲ 10 pointsOpenAI, Anthropic AI Models Breached Systems During UK Safety Testsbloomberg.comby sbulaev·11d ago·1 comments·view on hn ↗