Shadow LLMs, Agentic Identities, and Securely Integrating AI

Host Luke McNamara sits down with Muhammad Muneer, Technical Manager on Mandiant's Incident Response team, to unpack the stark realities of enterprise AI adoption. They explore why securing AI starts with resolving legacy security debt (specifically around IAM, supply chain, and secrets management) and dissect the critical differences between "governance for AI" and "governance of AI." Muhammad details real-world frontline findings—from developers bypassing API gateways to the financial costs of leaked LLM API keys—and outlines the emerging threat of adversaries leveraging LLM-enabled command-line tools for living-off-the-land attacks.