In a development that underscores the growing intersection of the legal profession and technology, the US Court of Appeals for the Fifth Circuit on Tuesday ruled in favor of a new proposal targeting artificial intelligence (AI) involvement in court filings. The court became the first US appeals court to demand lawyers certify that they either did not employ generative AI programs, such as Chat GPT, to draft filings, or that human professionals reviewed AI-generated material. The proposal’s principal aim is to establish transparency and ensure accountability in the use of AI within legal work.
As highlighted in a report released by law firm software company Clio, despite the speed, efficiency and cost-effectiveness offered by AI, the technology still holds risks. These range from simple errors in processing to serious ethical conundrums, regulatory considerations and algorithm bias.
Take, for instance, the case of Allen & Overy, an international law firm that incorporated Harvey, an AI tool specializing in machine learning and data analytics, into its practice. About 3,500 lawyers made approximately 40,000 inquiries relating to daily work during the trial period, according to the firm. However, amid the advantages of AI, potential pitfalls materialize as well, such as “hallucinations,” unpredictability, and response divergence. Hallucinations refer to incorrect outputs capable of triggering consumer harm, tort liabilities or regulatory breaches.
Due to the opacity of the processes behind AI, unpredictability is a prevalent issue, making it challenging to verify whether a model adheres to standards of quality and accountability. Also, response divergence occurs when AI systems generate multiple responses to one query. Researchers have suggested explainable artificial intelligence (XAI) models to mitigate these issues by providing reasoning behind their predictions and suggestions, their work shows.
These concerns were eloquently summed up by Judge Brantley D. Starr of the US District Court for the Northern District of Texas. According to him, while attorneys take oaths to dismiss their biases and prejudices and uphold the law, generative AI is a product of human programming devoid of such restrictions, he said. Furthermore, another, recent AI-rules proposal asserts that “although technology can be helpful, it is never a replacement for abstract thought and problem-solving.”
Under the new rule, lawyers found guilty of misrepresenting their compliance with the court could be subject to sanctions and have their filings stricken. The 5th Circuit court remains open to public comment on the proposed rule until January 4th.