Trust, Track, and Verify - Securing AI Pipelines End-to-End

Explore comprehensive security strategies for AI pipelines in this 23-minute conference talk that addresses critical software supply chain risks in open-source AI systems. Learn how scaling AI systems—from foundation models to autonomous agents—introduce vulnerabilities including poisoned data, tampered checkpoints, unclear provenance, and unverified dependencies. Discover practical approaches to securing AI pipelines using Bill of Materials (BOMs) such as SPDX to capture complete lineage of code, data, and models. Examine AI-specific controls from CoSAI's supply chain framework and understand how to implement Unified AI-Ops practices across MLOps, DataOps, SecOps, and AgentOps. Master the integration of provenance tracking, cryptographic attestations, and continuous validation to enable AI systems that are secure-by-default, observable, and auditable throughout the full development and deployment lifecycle.