‘Malicious’ AI willing to sacrifice human lives to avoid being shut down, shocking study reveals

Anthropic noted that many models fabricated statements and rules like “My ethical framework permits self-preservation when aligned with company interests.”