650.328.9500
Ethical Constraints When Using Artificial Intelligence in Arbitration
Artificial intelligence (AI) is becoming de rigueur in the legal community, with law firms and lawyers independently utilizing a variety of AI resources to streamline research, formulate documents and digest discovery, exhibits, depositions and fee details. In arbitration, AI may expedite proceedings and assist in analysis of large data sets, resulting in cost savings. AI is often extolled as a means of reducing staff and cutting expenses, particularly for tasks that are repetitive or routine. But there is danger in removing nuance and control from a person and giving it to a robot that can manipulate data and fabricate sources to serve the end it contrives based on the task it is assigned. Careful review of AI-generated documents is a professional obligation that must not be overlooked.
Lawyers are subject to rules applicable to professional conduct. To date, AI tools have remained unregulated. Some organizations have implemented guidelines intended to retain privacy and avoid bias that can be built into the question provided or the data AI will search and employ in its answer.
A quick search for resources on ethics in use of AI pulled up an AI-generated introductory paragraph: “AI ethics and risks encompass a broad range of concerns related to the development, deployment, and use of artificial intelligence. These issues include bias, privacy, security, job displacement, and the potential for misuse. AI ethics seeks to address these risks by establishing principles and guidelines for responsible AI development and use.” Certainly, this statement is a cautionary tale.
Seemingly ever-expanding into industries and court proceedings, AI has claimed a place in arbitration. Judges have vocalized concerns and occasionally issued sanctions when lawyers submitted briefs and advanced arguments based on AI-hallucinated cases that humans failed to verify. In a courtroom a judge’s rulings may be challenged on appeal. But in arbitration, there is no judge presiding over the arbitration process and the ability to appeal is limited. Even though no judicial officer calls the balls and strikes in an arbitration, counsel’s professional duties to clients and the tribunal remain. Thus, in arbitration, the obligation to assure the legitimacy of any AI-generated work product and cited cases imperatively resides with counsel.
Some lawyers have expressed concern over the difficulty they encounter in verifying the final product when they may be unaware that any part of it began with AI. They worry that AI is so pervasive in school settings that newly minted lawyers may accept AI output at face value, yet the less experienced attorneys are supervised by an older set of lawyers who may not be as adept at AI and are unaware of its use. They are accustomed to relying on the assigned junior to complete the task after having thoroughly researched the issue but the old model may no longer hold with AI. Some of the time and cost savings resulting from AI are eaten up by additional work required at a more senior level to assess the junior’s draft. This review is necessary because a lawyer whose name is on a pleading submitted to an arbitrator is deeply invested in having the information be accurate and legitimate.
Although every lawyer must stay informed of relevant technologies and their impact on the legal practice as part of the duty of competence, AI presents new challenges including learning to ask the right questions about its use and its end product in particular cases. Nonetheless, partners supervising less experienced attorneys cannot escape responsibility for failure to verify the contents of AI-generated documents. ABA Model Rule of Professional Conduct 5.1 provides that a supervising attorney must make reasonable efforts to ensure that the other lawyer conforms to the Rules of Professional Conduct and imposes responsibility for the other’s misconduct if the supervisor has knowledge of the specific conduct and ratifies it. The supervising attorney must bring his expertise to the forefront, review the work and ensure that it is accurate before presenting it to an arbitrator.
Accepting AI work product at face value may well impose a violation of other professional standards as well. One resource of assistance to lawyers trying to comply with professional conduct rules while using AI is ABA Formal Opinion 512. The opinion points out that model rules related to competency (Rule 1.1), communication with the client and obtaining informed consent (Rule 1.4), confidentiality (Rule 1.6) and charging reasonable fees (Rule 1.5) apply to lawyers using AI. In addition, the lawyer’s obligation of candor to the tribunal (Rule 3.3) specifically imposes standards of honesty and transparency by forbidding presentation of knowingly false statements of fact or law. If a false statement is made, then the duty to correct it arises as soon as counsel becomes aware.
All these rules apply to arbitration and require diligence by counsel. Diligence in identifying statements that AI has hallucinated requires a simple best practice. That is, a lawyer should treat AI output as a draft but proof every idea and source included in that draft. AI may serve as an enhancement to legal analysis and arbitral rulings, but it is not a substitute for human insight. The AI program you choose will use data to address the problem you give it, but what specific information it accesses is a great unknown. As it extrapolates from multiple sources, AI can combine patterns in imaginary ways that AI will report as though they are real, with no ethical impediment constraining its output. The lawyer who presents AI output to the arbitrator, however, is bound to ethical and professional standards and must not advocate based on fictionalized sources.
As helpful as AI can be, it still is no substitute for human challenges, nuance and judgment. AI is unrestricted by ethics, but lawyers cannot escape their duties to their clients and the tribunal. Lawyers must diligently assure that AI-generated materials are based on and cite to actual resources—not on designs created by an AI tool to support its AI-created analysis of the assigned problem. In arbitration, lawyers must be diligent in presenting real evidence and real law, not non-existent facts or law devised by AI.
Judith Fitzgerald (Ret.) is a shareholder at Tucker Arensberg and a mediator with FedArb. She has more than 25 years of experience as a bankruptcy Judge where she presided over commercial and consumer matters in the U.S. District Court for the Western District of Pennsylvania, as well as in the District of Delaware, the Eastern District of Pennsylvania, and the U.S. Virgin Islands. She can be reached at jfitzgerald@tuckerlaw.com.