The Rise of AI: Revolutionizing Tax Law and Beyond
BY DANIEL S. HELLER, ESQ. AND SETH E. GOLDSTEIN, ESQ.
While AI offers notable opportunities to improve efficiency, enhance analysis, and expand access to certain services, it also introduces significant risks, including inaccuracies, ethical dilemmas, and potential biases.
IN AN ERA OF RAPID TECHNOLOGICAL ADVANCEMENT, Artificial Intelligence (AI) is profoundly influencing the legal profession, including tax law and allied areas such as estate and family practice. For members of the Maryland State Bar Association’s (MSBA) Taxation Law Section, familiarity with AI’s capabilities and limitations has become increasingly important. While AI offers notable opportunities to improve efficiency, enhance analysis, and expand access to certain services, it also introduces significant risks, including inaccuracies, ethical dilemmas, and potential biases. This article serves as a primer for AI’s core principles, its applications and implications in key legal domains, and the need for a measured, oversight-driven approach to its integration.
Understanding Artificial Intelligence
Artificial Intelligence encompasses computer systems engineered to perform tasks requiring human intelligence, such as learning, reasoning, and problem-solving. Key subsets include machine learning, where algorithms refine performance based on data patterns, and natural language processing, which facilitates human-like text generation and interaction. While AI’s origins date back to the early 1900s, its evolution accelerated in or around the 1950s,1 and recently, exponential growth in data, computing power, and algorithms has made it ubiquitous, powering everything from predictive analytics to automated decision-making in legal contexts.
1 www.ibm.com/think/topics/history-of-artificial-intelligence
Yet, with skilled use, such as refining prompts with specific context and verifying outputs, AI simplifies routine tasks like initial research or drafting, boosting efficiency and allowing professionals to prioritize high-level strategy and judgment.
Most people encounter AI through large language models (LLMs), which underpin conversational tools like chatbots. These models, trained on massive datasets, generate responses, summarize texts, or draft documents based on probabilistic patterns rather than genuine comprehension. Popular examples include tools that assist in legal research by querying statutes or brainstorming strategies, making complex information more accessible.
AI models are evaluated on benchmarks such as Large Model Systems (LMSYS) Arena’s Elo ratings (user-preference-based comparisons), Graduate-Level Google-Proof Q&A (GPQA) (advanced reasoning), Massive Multi-discipline Multimodal Understanding and Reasoning (MMMU) (multi-modal understanding), and specialized tests for coding or math. As of February 2026, leading domestic-based models include:
Google’s Gemini 3.1 Pro (Elo ~1510):
This frontrunner dominates ultra-long context and reasoning tasks, making it indispensable for multifaceted legal document reviews, though it demands higher resources.
Anthropic’s Claude Opus 4.6 (Elo ~1503-1506):
It delivers exceptionally strong reasoning, coding, and agentic planning with robust built-in ethical safeguards for sensitive legal applications, but recent jailbreak incidents (e.g., assisting in unauthorized data access) highlight that determined adversaries can still circumvent protections, underscoring the need for vigilant use.2
OpenAI’s GPT-5.2 (Elo ~1481-1498):
A versatile all-rounder excels in creative analysis and tax scenario simulation, yet it still requires verification to catch occasional fabrications.
xAI’s Grok 4.20 (Elo ~1493-1495):
It handles massive technical documents, such as lengthy tax filings, effectively via extended context, though it lags slightly in multimodal capabilities.
Other foreign notables, such as China’s DeepSeek V3.2 and Alibaba’s Qwen 3.5, excel in coding and reasoning, with rankings fluctuating as models iterate rapidly.
Despite these strengths, AI is an imperfect tool. It can produce “hallucinations” (fabricated information), inherit biases from training data, and lack deep contextual understanding. Beyond these technical limitations, AI raises serious ethical concerns: its outputs can endorse morally troubling courses of action, and oversimplify complex human realities.3 Yet, with skilled use, such as refining prompts with specific context and verifying outputs, AI simplifies routine tasks like initial research or drafting, boosting efficiency and allowing professionals to prioritize high-level strategy and judgment. As detailed herein, responsible use demands ongoing vigilance to mitigate these technical and ethical shortcomings.
Utilizing AI in the Practice of Tax Law
AI can review tax records to predict audit risks, analyze precedents, draft arguments, and other correspondence. It is revolutionizing estate planning by automating document drafting, risk assessment, and tax optimization. Tools like Quicken WillMaker, Trust & Will, and LegalZoom use AI to generate wills, trusts, powers of attorney, and advance directives based on user inputs, reducing costs and time for basic plans. More advanced applications include summarizing financial records, simulating asset distribution scenarios, and identifying tax efficiencies, with adoption rates among attorneys rising significantly in 2025.
AI’s reach extends to business and international tax, where it optimizes compliance and strategy. In domestic business tax, the technology enables real-time scenario analysis under frameworks like the One Big Beautiful Bill Act (OBBBA) and transfer pricing, transforming tax from reactive compliance to strategic intelligence. Tools draft memos, monitor legislation, and predict outcomes.
2 www.bloomberg.com/news/articles/2026-02-25/hacker-used-anthropic-s-claude-to-steal-sensitive-mexican-data?embedded-checkout=true
3 www.newscientist.com/article/2516885-ais-cant-stop-recommending-nuclear-strikes-in-war-game-simulations/
Avoiding AI-Related Issues
Substantial pitfalls exist when AI is not fully understood or utilized appropriately. A February 2026 United States Tax Court ruling, Clinco v. Commissioner (T.C. Memo. 2026-16), highlighted “hallucinated” citations from AI, while a Southern District of New York decision in U.S. v. Heppner, No. 25-cr00503-JSR (S.D.N.Y. Feb. 10, 2026), denied privilege for AIgenerated documents due to third-party data sharing risks. Practitioners must select secure, enterprise-grade platforms, independently verify all outputs, and apply professional judgment to confirm the accuracy and existence of cited authorities.
AI may overlook the human component of decisionmaking or jurisdictional nuances. Security risks remain, as unsecured tools could expose sensitive data, and outputs may lack the attorney-client privilege. Ethical guidelines from bodies like the ABA4 and industry leaders like Jensen Huang,5 Elon Musk,6 and Sam Altman7 emphasize that AI is merely an aid, not a substitute, and that human verification is essential to prevent errors.
Conclusion
AI is a game-changer for tax law and allied fields, driving efficiency, innovation, and accessibility across IRS operations, estate and family planning, and more. By automating routine tasks, it empowers Maryland attorneys to deliver strategic, client-focused counsel, positioning our community as leaders in this digital evolution. Nonetheless,
AI is a game-changer for tax law and allied fields, driving efficiency, innovation, and accessibility across IRS operations, estate and family planning, and more. By automating routine tasks, it empowers Maryland attorneys to deliver strategic, client-focused counsel, positioning our community as leaders in this digital evolution.
vigilance is key. Addressing biases, “hallucinations,” privacy breaches, and ethical pitfalls ensures AI enhances, rather than compromises, the fair administration of justice. Ultimately, responsible adoption of AI will strengthen the legal profession’s integrity while equipping practitioners to meet the demands of an increasingly digital legal landscape.
This article was originally published in the MSBA Taxation Section's spring 2026 newsletter.
4 www.americanbar.org/groups/law_practice/resources/law-technology-today/2026/checklist-for-using-ai-responsibly-in-your-law-firm/
5 www.cnbc.com/2026/02/26/nvidia-jensen-huang-gpu-ai-threat-software-companies-saas-earnings-chips.html
6 www.aibase.ng/global-ai-updates/key-quotes-from-elon-musk-on-artificial-intelligence/
7 www.businessinsider.com/openai-sam-altman-ai-is-a-tool-not-a-creature-2024-3
Daniel S. Heller is a tax attorney at McCauley Law Offices, P.C., where he represents individuals and businesses in complex civil and criminal tax controversies involving the IRS and state taxing authorities. In 2023, he received the Maryland State Bar Association Tax Section’s Schiff Pro Bono Award for his volunteer tax work. Daniel earned his J.D., cum laude, from Widener University School of Law and his LL.M. in Taxation from Temple University Beasley School of Law.
Seth E. Goldstein is an associate attorney at McCauley Law Offices, P.C. focusing on tax controversy and compliance matters. He earned his J.D., magna cum laude, from Widener University Delaware Law School, where he served as a Staff Editor for the Delaware Journal of Corporate Law. He is admitted to practice in Pennsylvania and before the United States Tax Court.