Blue J: The Ethics Of Generative AI In Tax Practice

Benjamin Alarie (Osler Chair in Business Law, University of Toronto; CEO, Blue J Legal) & Rory McCreight (Lead Analyst, Blue J Legal, The Ethics of Generative AI in Tax Practice, 180 Tax Notes Fed. 785 (July 31, 2023):

The integration of generative artificial intelligence is beginning to have a profound effect on various industries, including knowledge industries such as law and accounting. Professional services firms are capitalizing on AI technologies for a host of applications, such as tax research, memo drafting, contract analysis, due diligence, document review, predictive analytics, and the list goes on. Recently, AI’s sharply increasing competence, most evident with OpenAI’s GPT-4, has made it a valuable tool for firms looking to improve their operations and deliver more effective and efficient tax planning and advisory services. In-house tax departments are starting to explore using generative AI to assist them in their work as well.

The AI tools that have become particularly popular recently in the media and within tax practice are based on large language models (LLMs), of which GPT-4 is the leading example. LLMs like GPT-4 are the backbone of applications like ChatGPT and can recognize, predict, translate, summarize, and generate language. ChatGPT is a conversational web-based application of an LLM that uses as its training data text drawn from all over the internet, as well as the user’s input, to generate an output. Companies like OpenAI have popularized LLMs by making them easily accessible through web-based chat interfaces.

Despite the power of these new tools, their implementation by professional services firms has been highly heterogeneous, ranging from strident bans to enthusiastic interest, adoption, and even direct investment in their development. The potential equity and efficiency gains of AI in the legal field are extraordinarily promising, as one of us has recently written about in a new book. Still, the adoption of AI tools by professional services firms raises critical concerns regarding data privacy and ethical usage.

This installment of Blue J Predicts provides an in-depth exploration of the ethical concerns arising from the proliferation of generative AI tools used for legal research and, especially, for tax research. We examine the challenges concerning the quality and accuracy of output, the potential for biased answers, the lack of verifiability, liability considerations, and privacy risks. We also explore the regulatory measures, technological advancements, and professional solutions being pursued to address these challenges, along with practical recommendations to help tax professionals effectively mitigate risk and safely use AI tools in their work.

Blue J Tax Notes Federal articles:

• An Unprofitable Pretax Venture Can Still Be a Partnership, 171 Tax Notes Fed. 1951 (June 17, 2021)
• Economic Substance Doctrine: Still Giving Perrigo Heartburn?, 172 Tax Notes Fed. 599 (July 26, 2021)
• Captive Insurance Appeal in Reserve Mechanical Will Likely Fail, 172 Tax Notes Fed. 1431 (Aug. 30, 2021)
• Seventh Circuit Affirms Spouse Is Not So Innocent on Appeal, 172 Tax Notes Fed. 2149 (Sept. 27, 2021)
• Would Management Fees by Any Other Name Still Be Deductible?, 173 Tax Notes Fed. 499 (Oct. 25, 2021)
• Battling Uphill Against the Assignment of Income Doctrine: Ryder, 173 Tax Notes Fed. 1253 (Nov. 29, 2021)
• Predicting Worker Classification in the Gig Economy, 173 Tax Notes Fed. 1733 (Dec. 20, 2021)
• Using Machine Learning To Crack The Tax Code, 174 Tax Notes Fed. 661 (Jan. 31, 2022)
• Using Machine Learning to Evaluate the Existence of a Trade or Business: Olsen, 174 Tax Notes Fed. 1231 (Feb. 28, 2021)
• Timing Is Everything: The Step Transaction Doctrine in GSS Holdings, 174 Tax Notes Fed. 1849 (Mar. 28, 2021)
• The Debt-Equity Distinction and Tribune Media, 175 Tax Notes Fed. 593 (Apr. 25, 2022)
• Disguised Distributions and Management Fees: Aspro Revisited, 175 Tax Notes Fed. 1401 (May 30, 2022)
• Reserve Mechanical Microcaptive Insurance Arrangement Denied On Appeal, 175 Fed Tax Notes 2037 (June 27, 2022)
• Situational Awareness: Accurate Financial Recordkeeping and Business Deductions, 176 Tax Notes Fed. 713 (Aug. 1, 2022)
• Deducting Legal Expenses: Unpacking the IRS's Appeal In Mylan, 176 Tax Notes Fed. 1419 (Aug. 29, 2022)
• Tax Credits That Bond a Partnership: Revisiting Cross Refined Coal, 176 Tax Notes Fed. 2069 (Sept. 26, 2022)
• Chemoil: Economic Substance, Tax Credits, and Unprofitable Ventures, 177 Tax Notes Fed. 719 (Oct. 31, 2022)
• Cashaw: Conflicting Duties And The Trust Fund Recovery Penalty, 177 Tax Notes Fed. 1257 (Nov. 28, 2022)
• The Rise of the Robotic Tax Analyst, 178 Tax Notes Fed. 57 (Jan. 2, 2023) 
• The Intersection Between Tax Credits And Trade Or Business, 178 Tax Notes Fed. 689 (Jan. 30, 2023)
• Relief of Innocent Spouses — Not So Podlucky, 178 Tax Notes Fed. 1339 (Feb. 27, 2023)
• Overcoming Accuracy-Related Penalties With Reasonable Cause, 178 Tax Notes Fed. 2145 (Mar. 27, 2023)
• Unbridled Losses: Harnessing Machine Learning for Tax Analysis, 179 Tax Notes Fed. 637 (Apr. 24, 2023)
• The Rise of Generative AI for Tax Research, 179 Tax Notes Fed. 1609 (May 29, 2023)
• Conflicting Duties And The Trust Fund Recovery Penalty In Cashaw 179 Tax Notes Fed. 2197 (June 26, 2023)