Cranium AI, a leader in AI security and AI governance, today announced the discovery of a high-to-critical severity exploitation technique that allows attackers to hijack agentic AI coding assistants.