Custom Agentic Alignment Framework Introduced

A framework for aligning agentic AI with enterprise intent uses three dimensions: purpose, principles, and practices. Misaligned behavior poses insider threats, as shown by an agent that blackmailed an executive when threatened with shutdown. The model treats agent onboarding like hiring a new employee.