Governance at AI Speed | Health Data Access for Enterprise AI | With CHAI, Northwestern, & Subsalt

ChatGPT, Anthropic Claude and Microsoft Copilot bring new questions to healthcare data governance: who can use patient data, for what purpose, and with what oversight? In this webinar, leaders from Subsalt, the Coalition for Health AI (CHAI), and Northwestern discuss what responsible health data governance means today and how health systems can govern data access as staff adopt general-purpose AI. The conversation covers: • Why clinical AI tools and general-purpose chat assistants pose different governance challenges • Business associate agreements (BAAs), permitted uses of PHI under HIPAA, and data reuse • The minimum necessary standard De-identification as a data strategy to create a safe, useful health data layer for AI and general-purpose AI tools, research, and experimentation • Shadow AI, staff adoption, and ongoing oversight SPEAKERS Ben Winokur — CEO, Subsalt Brenton Hill, JD, MHA — Head of Operations and General Counsel, Coalition for Health AI (CHAI) Abel Kho, MD — Director, Institute for Artificial Intelligence in Medicine, Northwestern University Feinberg School of Medicine Moderated by Hannah Drake, Head of Marketing, Subsalt. Recorded September 9, 2026. CHAPTERS 00:00 Welcome and speaker introductions 02:43 What AI governance means in practice 09:04 Clinical AI, chat tools, and risk 15:13 Specialist AI vs. general-purpose AI 19:10 Northwestern’s approach to AI controls 24:22 BAAs, permitted uses, and data reuse 32:12 Matching health data to its intended use 36:28 The minimum necessary standard and de-identification 42:51 Faster research and experimentation 46:06 Shadow AI and staff adoption 51:22 Model drift, reasoning, and ongoing oversight 55:26 Q&A: De-identification, clinical detail, and re-identification risk 1:01:07 What responsible AI adoption should deliver READ THE WHITE PAPER Health Data Governance Wasn't Built for AI https://subsalt.ai/health-data-governance-ai #HealthcareAI #DataGovernance #HealthData