Healthcare organizations are moving quickly to adopt AI-enabled tools, including scheduling agents, website chatbots, and ambient scribes. As they do, the contracts behind these tools are getting longer, more complicated, and more contested. Customers now routinely attach their own terms on data privacy, intellectual property, and AI use, which can override a vendor’s standard agreement. Vendors, in turn, have to work out whether they can comply with myriad contractual restrictions while simultaneously protecting their own IP and negotiating the provisions that allow their tool to provide the desired services.
In the accompanying video, Nixon Peabody partner Valerie Montague and counsel Jason Kunze are joined by Shawna Doran, Vice President, Associate General Counsel and Chief Privacy and Security Counsel at Veradigm. Together, they look at how HIPAA and other privacy regulations, intellectual property, confidentiality, and AI-specific contract terms are coming together in healthcare technology contracts, and what that means for providers, health plans, life sciences companies, and the vendors that serve them.
Privacy and data protection
- Map the rules before you negotiate. Besides HIPAA, organizations may be subject to 42 CFR Part 2, state health privacy laws protecting specific types of data (such as mental health, HIV/AIDS, genetic, biometric, and reproductive health information), state comprehensive privacy laws, and new state AI and health AI laws. Board policies, offshoring restrictions in other contracts, and public privacy commitments may also limit how data can be used or shared with a vendor. It’s also important to confirm what data elements are actually involved in a proposed arrangement. For example, a HIPAA-regulated organization may be buying a tool that doesn’t use PHI, and whether data counts as de-identified depends on the HIPAA and state law standards that apply.
- Clarify the vendor’s role and data access. Organizations need to establish whether a vendor is a business associate or a subcontractor business associate, whether it needs the full record or only a limited subset, consistent with the HIPAA minimum necessary requirement, and whether existing EMR and other vendor agreements permit integration, to the extent applicable. For example, an AI scheduling agent that accesses the EMR may not need access to the full medical record. Subcontractor flow-downs add another layer, particularly when a vendor’s subcontracts are already in place and may not be able to be easily modified. These questions, along with the vendor’s security controls, are best addressed through diligence and governance before contracting begins, especially with newer companies that may have limited experience handling health information.
- Ambient scribes and AI note takers raise distinct issues. These include patient consent in two-party consent states, whether recordings and transcripts can be obtained in litigation, how long they’re kept, and whether they form part of the medical record. Recent lawsuits over AI note-taking tools have included biometric privacy claims and whether such a tool acts for a participant in the conversation or counts as a third party to it. How contracts allocate responsibility for consent and note review is an important issue for both providers and vendors.
- Patient-facing chatbots require careful design. Some states have adopted or are considering laws on AI and the practice of medicine. Users also often volunteer health information in chatbot conversations. Organizations deploying these tools need to consider how a chatbot handles that information and when it should direct users to clinical care.
Intellectual property and data rights
IP ownership and license scope remain core terms. For vendors, customer terms that alter IP ownership or restrict the right to modify the software or create derivative works are often among the most important points to negotiate. License scope matters on the inbound side, too. Third-party data licensed for limited use may prohibit AI training or use, and the data provider may seek additional fees if AI use extends beyond the original license.
Data reuse rights carry both value and risk. Whether a vendor may use customer data, including de-identified data and retained prompts, to train or improve AI models is often a central negotiating point and can affect deal economics. Linking de-identified datasets doesn’t necessarily produce a de-identified result, and AI can increase the risk of re-identification. That makes downstream contract protections especially important for organizations licensing health data.
AI-specific contract terms
- AI riders may take precedence over the main agreement. Customers often attach AI riders or addenda that purport to control over the vendor’s standard terms. These provisions typically shift more risk to the vendor, for example by adding warranties regarding AI usage, removing liability caps, and increasing insurance requirements.
- Broad AI definitions can reach beyond their intended scope. Some contracts define AI so broadly that the definition covers nearly any software. When that definition triggers notice, approval, or restrictions on use, it can reach routine work like support and testing. Refining definitions to match the use cases and potential risks is critical to reaching a workable agreement.
- Confidentiality commitments should match the AI supply chain. Many AI tools, particularly public or free versions, do not protect the confidentiality of the data entered into them, and data used to train a model can be hard to remove. In recent protective order and privilege disputes, courts have required private or enterprise AI environments with enforceable confidentiality terms. In one case, a court found that a provider’s terms allowing it to share data with third parties were enough to defeat a privilege claim. A vendor’s promises to its customers are only as strong as its own agreements with AI companies and subcontractors.
As health technology becomes more central to care delivery and operations, contract terms increasingly determine how organizations manage regulatory exposure, data rights, and liability. Nixon Peabody’s healthcare, privacy, and intellectual property attorneys advise healthcare organizations and the technology vendors that serve them on these arrangements.

