Legal Experts Advise Hybrid Contracts for Agentic AI
As AI products become more autonomous, legal experts recommend hybrid contracts that blend SaaS terms with service-level agreements typical of business process outsourcing (BPO). This new model suggests including business outcome guarantees and risk-sharing clauses. The shift is particularly relevant for productized services where AI tools take on active decision-making roles that still require expert oversight.
- Under traditional Software-as-a-Service (SaaS) agreements, the customer often bears the risk for actions taken by an AI agent, with suppliers providing the technology "as is" and disclaiming responsibility for accuracy or fitness for purpose. These contracts frequently exclude liability for the very harms agentic AI is most likely to cause, such as lost profits, regulatory fines, or reputational damage. - The proposed hybrid model redefines key contractual clauses, such as Service Level Agreements (SLAs). Instead of measuring only technical uptime (e.g., 99.99%), the new SLAs are operational and tied to business outcomes, such as the accuracy of tasks performed by the AI agent. - Data and intellectual property rights are a central point of negotiation. While some SaaS terms grant the provider broad licenses to use customer data for model training, the hybrid BPO-style approach typically stipulates that the customer owns both the inputs and the outputs of the AI service, requiring explicit consent for any data to be used for training. - The EU's AI Act
