AI Contract Drafting 101: Model Governance Provisions Explained

Laura Belmont (GC at The L Suite), Matt Kohel (Partner at Saul Ewing), and Laura Frederick (Founder of How to Contract) discuss core concepts for drafting AI model governance provisions in vendor contracts. In this conversation, they cover: - Why AI model provisions differ from traditional software contracting - The flow-down limitations between frontier model providers, vendors, and customers - What vendors can realistically commit to (and what they cannot) - The "black box" challenge and proprietary constraints around training data and weighting - Why models are dynamic and what that means for static contract language - Vendor concerns around broad cooperation obligations and undefined scope - What customers should push for: data training, retention, performance thresholds, and notice of material changes - Why cooperation provisions matter more than full transparency - The case for replacing "reasonable" with "thoughtful" in AI contracts This session offers practical guidance for in-house counsel, outside counsel, contract managers, and procurement professionals working on AI vendor agreements. 00:00 Why AI model provisions matter 00:30 The three-party framework (customer, vendor, model provider) 02:00 The proprietary "black box" challenge 03:30 Why AI is dynamic, not static 04:30 Vendor tensions with broad cooperation obligations 06:30 What vendors can reasonably offer 08:00 Customer perspective: what to actually push for 10:00 Data training, retention, and material change notices 11:30 Why cooperation provisions matter #AIContracts #ContractDrafting #VendorContracts #AIGovernance #LegalTech #InHouseCounsel #HowToContract