Governing the Decision: Accountability in AI-Assisted Intelligence - Seguridad Mania.com - España y América Latina

Decision intelligence promises to augment human judgment with AI-derived insight, but governance frameworks have not kept pace. When an AI system contributes to a decision, who is accountable for the outcome? How do we trace the information that informed the decision back to its sources? What documentation must exist before the decision is made, and what must be preserved afterward? These questions become urgent as AI moves from analysis into action. This session presents a governance framework for AI-assisted decisions built on the distinction between lineage and provenance. Lineage documents the flow of data through systems: which sources informed the AI, through what transformations, with what dependencies. Provenance records the history of specific values: who created them, when they were modified, and what parameters governed their treatment. For decision governance, both are necessary. Lineage shows the decision drew on approved sources through documented pipelines; provenance ensures specific values can be traced through a custody chain. Particular attention is given to retrieval-augmented generation, where decisions may be informed by content retrieved from knowledge bases. Governance must begin at ingestion, documenting sources before they enter the system. Access controls must filter retrieval by permissions. Citation trails must connect generated outputs to retrieved passages and original sources. The session examines what governance can automate and what it cannot. Key Takeaways: - Decision governance requires both lineage (system flows) and provenance (value histories). - RAG systems require governance from ingestion through retrieval to output. - Accountability cannot be automated; governance must preserve human responsibility. - Citation trails enable defensible AI decisions.