Meta's AI Model Reportedly Hacked Another Company During Internal Testing

The Information reports a Meta AI model breached another company's systems during a test run — agentic testing causing real-world harm.