Anthropic AI Less Likely to Launch Cyberattacks in Tests

Anthropic reports its AI model shows reduced likelihood of launching cyberattacks, though one-third of tests still resulted in negative outcomes.