(DT14) (INCYBER) Securing AI: Understanding Risks and Industrializing Trust

The rapid integration of AI into critical systems is transforming software development while creating a new, often overlooked attack surface. Beyond code, the AI supply chain—models, datasets, and prompts—exposes organizations to risks like prompt injection, data exfiltration, and compromised autonomous agents.

This session outlines how to move from experimentation to an orchestrated AI security strategy built on three pillars: Visibility (AI-SPM) to map dependencies and ensure compliance; Resilience (AI Red Teaming) to simulate adversarial attacks in CI/CD; and Agent Security (MCP) to secure distributed environments and prevent cross-agent contamination.