Public to Proprietary: Legal Strategies for AI Data and Model Governance

As artificial intelligence transitions from experimentation to enterprise use, legal and brand risks increasingly depend on the underlying data and models. Open datasets and open-source models seek to democratize AI but open-source models shift the legal risk primarily to the end users. Open datasets can be messy and often combine smaller sets with unique license terms, which can introduce obligations and risk into downstream models. 

This panel examines AI risk across the lifecycle — from training data for purpose-built models to deployment of open-source models — through a practical, business-focused lens. Panelists will explore how composite datasets, open-source and open-weight models, and third-party components layer legal, IP, licensing and privacy issues. The discussion will also address real-world risk realization and how litigation and enforcement realities shape defensible risk decisions. Attendees will leave with a practical, risk-based framework to guide governance, document tradeoffs and support innovation without creating avoidable downstream exposure.

What you will learn:

• Methods for evaluating tangible legal risks in the use of open-source and open-weight models. 
• Real world approaches to managing open-source, copyright and privacy obligations in training datasets. 
• A practical, jurisdiction-aware risk framework to guide AI business decisions.