The Rise of Autonomous Artificial Intelligence in Contract Negotiation
Robert Handler
In the final months of 2023, an artificial intelligence successfully negotiated a contract with another artificial intelligence completely autonomously.[1] With AI assistance, a contract can go from a blank document to a complete contract ready for signatures in a few minutes.[2] This remarkable feat was accomplished by Luminance, an AI developer based in the UK, by developing a revolutionary large language model to independently analyze and alter contract terms.[3] On a general level, the software works by ranking the most contentious terms or clauses, and then prioritizing which are most contentious to least.[4] The contentious clauses are then changed to something more suitable, while the program records the alterations for later review.[5]
Since this initial breakthrough, some of the world’s largest companies, such as Walmart, Maersk, and Vodaphone, have elected to use AI to negotiate and maintain their own supplier contracts.[6] This use of AI allows these companies to almost instantaneously develop contracts for the thousands of small suppliers that are vital to their business. However, on some occasions, this AI model has delivered some creative results.[7] In one instance, a prominent AI researcher allowed two AI bots to negotiate over the hypothetical premise of where to order dinner.[8] One bot was instructed to want pizza for dinner, while the other bot was coached to want sushi.[9] After a “successful” negotiation, the bots agreed to put sushi on the pizza.[10] In addition, a Polish radio station came under fire after firing their on-air host and replacing them with an AI “chatbot”. This AI chatbot decided to host an interview with Wislawa Szymborska, the winner of the 1996 Nobel Prize for Literature.[11] However, Wislawa Szymborska had been dead for 12 years prior to the interview.[12] The entire interview was a fabrication of the AI chatbot.[13]
This technology, although clearly in its infancy, could have widespread ramifications over the legal field, business efficiency, and how consumers interact with corporate interests. These ramifications may have longstanding effects on the relative power dynamics of a negotiation. In traditional negotiations, inefficiency is somewhat necessary due primarily to information insecurity and advantageous self-serving.[14] In other words, one side of a negotiation may believe it advantageous to misrepresent their position, in the hopes that the opposing side does not have information to the contrary, or will bargain based on this false position to the benefit of the lying party.[15] In the future, AI negotiators will be able to utilize algorithms that deduce interests from behavioral patterns, assess real-time information derived from verbal and non-verbal communication, and profile human negotiators.[16] In addition, AI negotiators will be able to comb through all publicly available information instantaneously and likely will be able to determine if positions are misrepresented.[17] While this may initially appear to be beneficial, and likely will have a positive impact on many aspects of a negotiation, there is also skepticism by many in the AI and legal communities. The capacity for manipulation will likely be reduced while the ability to negotiate an agreement that best matches each party’s respective preferences will likely increase.[18] Based on these predictions, it appears as though the AI negotiating future is bright; decreased cost, emotional labor, and time, all while fair distributions and predictable uniform results are maximized[JS1] .[19]
However, possibly the most important question remains: Is it right to remove the humanity from negotiation? As mentioned previously, current AI models may still create solutions that are opposed to human sensibilities. While this is a significant hurdle to overcome, it is likely a limitation of current AI models that will improve over future generations. Another obstacle facing AI contract negotiators revolves around the limitation of non-human machines. Some experts in the field of artificial intelligence have cautioned that intention, an essential prerequisite of legally enforceable contracts, can never be generated by an artificial intelligence.[20] Without outwardly expressed intention, an entirely human construction, the validity of AI negotiated contracts is called into question under the current theories of contract law.[21] In response, however, supporters of AI negotiated contracts contend that if an AI can be developed to mimic the outward expression of intent that is indistinguishable from a human to the reasonable observer, it should have no ramifications on the legal system.[22] While this opinion corresponds with current thought regarding AI intelligence measurements, its implementation in the United States legal system may not be smooth.[23] Regardless, until AI can be trusted to run completely autonomously and operate indistinguishably from humans in contract negotiations questions remain regarding the intention of these agreements.[24]
[1] Kif Leswing, AI Negotiates Legal Contract Without Humans Involved for First Time, CNBC (Nov. 12, 2023, 10:56 am), https://www.cnbc.com/2023/11/07/ai-negotiates-legal-contract-without-humans-involved-for-first-time.html.
[2] Id.
[3] Id.
[4] Id.
[5] Id.
[6] Michael Dumiak, AI Agents Are Taking Over Contract Negotiations, IEEE Spectrum (Nov. 7, 2024), https://spectrum.ieee.org/ai-contracts.
[7] Id.
[8] Id.
[9] Id.
[10] Id.
[11] Andrew Higgins, Polish Radio Station Uses A.I. to Interview Dead Nobel Laureate, N.Y. Times (Nov. 3, 2024), https://www.nytimes.com/2024/11/03/world/europe/poland-radio-station-ai.html.
[12] Id.
[13] Id.
[14] Horst Eidenmüller, Game Over: Facing the AI Negotiator, U. Chi. L. Rev. Online 1, 2–3 (2024).
[15] Id.
[16] Id. at 3.
[17] Id.
[18] Id. at 5-7.
[19] Id.
[20] Eliza Mik, AI in Negotiating and Entering into Contracts, in The Cambridge Handbook of Artificial Intelligence: Global Perspectives on Law and Ethics 45, 45 (Mark Findlay & Jolyon Ford eds., 2024).
[21] Id. at 48.
[22] Id. at 50-51.
[23] Id.; See also, The Turing Test, Stanford Encyclopedia of Philosophy, (Oct. 4, 2021), https://plato.stanford.edu/entries/turing-test/.
[24] Eliza Mik, AI in Negotiating and Entering into Contracts, in The Cambridge Handbook of Artificial Intelligence: Global Perspectives on Law and Ethics 45, 48 (Mark Findlay & Jolyon Ford eds., 2024).
[JS1]Consider adding a citation here