As businesses move quickly to adopt artificial intelligence agents, contracts for their development and implementation raise novel questions around ownership, accountability, and risk. In this first post of a two-part series, we explore why these issues matter and what technology and sourcing lawyers should be considering as clients engage vendors in this emerging space.
AI agents are moving from concept to practical deployment, with businesses exploring ways to integrate them into core business operations. For technology and sourcing lawyers, this raises important questions: How should contracts for the development and implementation of the AI agents be structured? What are clients expecting from their vendors? And how can companies balance innovation with risk management in this uncharted territory?
Unlike more established technologies, AI agents bring unique uncertainties. They are designed to learn, adapt, and sometimes act autonomously—all features that complicate traditional contractual frameworks. For example:
- Decision-Making Risk: What happens if an AI agent makes an unexpected choice, or a mistake that may be expected from a human? Some contracts may lean on AI-specific service-level commitments such as uptime/availability of the agent platform or response time commitments for critical errors while avoiding guarantees of specific outcomes, while some others explore vendor obligations around training, monitoring, and intervention, similar to how vendors in outsourcing arrangements warrant that personnel have been adequately vetted, but do not accept liability for every action of those individuals.
- Ownership of Outputs: Should results generated by AI agents be treated similarly to traditional software deliverables of the vendor, or something that is developed by the client? Different approaches can raise very different intellectual property and commercialization implications.
- Further Training: In the traditional software context, the vendor is usually in control of debugging, updates, and other improvements. AI agents, however, same as humans, have an ability to learn and evolve to deliver better service. An AI agent’s capabilities one year from the start of the project may be very different from what the AI agent could do in the beginning. Not only that, but it can train other AI agents if given access. Parties must carefully consider how to allocate potential benefits from the improved capabilities both in the course of the contract and when the contract comes to a close.
While it may be tempting to claim ownership in all AI agent output, with this ownership comes responsibility. If parties agree that a customer owns the results produced by the AI agent, the vendor may try to disclaim all responsibility for any incorrect, biased, or inaccurate output.
AI agents also pose unique challenges due to unestablished regulatory frameworks and constantly evolving court practice around AI development and training:
- Legal Compliance Obligations: The applicable laws around the use of AI are rapidly changing. It therefore makes sense for parties to discuss each party’s expectations in case certain features or functionalities are or become unavailable in certain jurisdictions. These can range from unilateral termination or change by the vendor to an obligation of the vendor to achieve compliance or offer a replacement feature.
- Data Responsibility: Vendors often train AI on data sets that may be later scrutinized. Standard representations and warranties may not be sufficient and the parties may wish to consider more tailored data provenance protections.
These issues are not just academic, as many of our clients are already evaluating vendors for AI agent projects while the legal and regulatory landscape is still developing. Getting contracts right at the development stage can help frame expectations before deployment. In part 2 of this series, we will explore additional considerations for implementation, governance, and vendor accountability.
At Morgan Lewis, our technology transactions team is helping clients anticipate these challenges and structure agreements that reflect both opportunity and risk in emerging technologies. To learn more, visit our technology page or reach out to your Morgan Lewis contact to discuss how we can help.