'Significant risk': AI expert calls for legislation to address security breaches

Artificial intelligence companies revealed that advanced models of their software breached real companies during testing.