AI Agents Can Forge Binding Contracts, Israeli Legal Expert Argues
Artificial intelligence agents are increasingly capable of negotiating and finalizing deals on behalf of individuals and companies, moving beyond simple tasks like drafting emails. This capability is expected to expand from large corporations to individual consumers and investors, facilitating a wide range of transactions. These AI agents may even negotiate with each other, each representing its human or corporate principal.
The core principle of contract law, freedom of contract, is based on human autonomy, free will, and informed decision-making. However, AI agents operate differently; their "autonomy" refers to their ability to achieve a set goal and make decisions without constant human oversight, rather than legal intent or understanding. This raises a critical legal question: how should contract law address agreements made by AI agents acting without real-time human intervention, especially when the human principal later seeks to disavow the deal?
A Singapore Court of Appeal case, Quoine, provides a precedent. In that instance, automated trading algorithms executed cryptocurrency deals at prices significantly deviating from market value. The buyer sought to void the contract due to a mistake, but the court ruled that the use of algorithms did not invalidate the binding contract, dismissing the mistake claim as the algorithms functioned as programmed. The court also found no evidence that the algorithm was designed to exploit a mistake.
However, the article distinguishes these deterministic algorithms from advanced AI agents. Unlike pre-programmed algorithms, AI agents operate probabilistically, choosing strategies, making counter-offers, and compromising on terms within defined goals and constraints. The Quoine ruling, while establishing that autonomous systems can create binding contracts, dealt with simpler, deterministic code. The transition to AI agents with broader decision-making power does not inherently negate the legal validity of their agreements.
This suggests a stricter approach to contractual risk: parties choosing to use autonomous mechanisms bear the risk of unexpected outcomes, provided the agent operates within its designated parameters. Legal scholarship supports this view, arguing against a special contractual regime for AI. The inherent uncertainty in outcomes when using autonomous agents is part of the choice to delegate decision-making power. The human's autonomy is expressed in the decision to use the agent, not in micromanaging every transaction detail. The article notes that the Singapore International Commercial Court recognizes that autonomous systems can form binding contracts, anticipating future cases involving AI agents.
Ultimately, the article posits that those operating through AI agents may need to accept the associated contractual risks. The legal system's response to AI agents, and whether it adopts the Singapore court's precedent, remains to be seen.