"As AI develops, competence may include the skills needed to use AI appropriately, depending on your job," said Adrienne Griffin of the IRS Office of Chief Counsel. "That's one question I've definitely been thinking about." (iStock.com/Wanniwat Roumruk)
The growing potential for using AI to respond to changes in federal tax regulation highlights the evolving role of tax attorneys. They must understand complex tax laws while also integrating swiftly developing AI tools into their practice in ways that ensure accuracy, compliance and ethical use, ultimately reshaping the traditional practice of tax law.
James McPhillips, a partner in Clifford Chance's global tech group, said lawyers who use the latest generative AI technology should always apply independent judgment, legal training and experience to AI-produced output before sending that work product back to their client.
"In addition, lawyers who use AI systems need to be aware how those systems are architected," that is, "whether the AI model is a public version or a version that is ring-fenced with enterprise security and controls," McPhillips told Law360. "If a lawyer introduces sensitive client information into a public AI model," such as the public subscription version of ChatGPT, "he or she could be exposed to the risk of breaching his or her duty of confidentiality to the client."
Adrienne Griffin, of the Internal Revenue Service Office of Chief Counsel, said tax lawyers have a professional duty to their clients to maintain their competence and monitor changes in tax law, but also to keep abreast of the benefits and risks associated with changes in AI technology. The development and use of AI could become a core skill in the legal profession, she said, speaking recently at the American Bar Association tax section's May meeting in Washington, D.C.
"As AI develops, competence may include the skills needed to use AI appropriately, depending on your job," Griffin said. "That's one question I've definitely been thinking about."
Other ethical concerns for lawyers using AI, as spelled out in the ABA's rules of professional responsibility, are requirements for lawyers to be able to competently and diligently determine the legal issues at stake, she said.
"Don't use AI as a complete shortcut or a way to procrastinate until the end," she warned. "You have to act with commitment and dedication to your client's interests."
AI can pull information from the website of one client and generate a letter that can be repurposed for another client, warned Macdonald Norman of Todd Welty PC, but that leads to some very obvious problems.
"You can't just copy and paste, you have to check to make sure the law is still correct. You have to make sure the facts are applied appropriately," Norman said. Eventually, AI might be able to do that type of comprehensive analysis, but "it's not coming for our jobs anytime soon," Norman said in his remarks before the ABA panel.
He noted that one of the biggest hurdles facing lawyers who use AI is determining responsibility.
"One of the big questions is the question of accountability," Norman said. "When something goes wrong with AI, what human is ultimately responsible for that?"
Alexandra Minkovich, a partner at Baker McKenzie, said AI could perform some basic tasks, such as providing a list of potentially important topics in proposed tax regulations that shouldn't be overlooked in comment letters.
AI can also make some general recommendations for improving a proposed regulation, but they might not be very significant, and they also might be areas that are outside the statutory authority of the U.S. Treasury Department or the IRS, Minkovich said at the ABA panel.
Artificial intelligence can "kind of save me the time of starting from scratch,"she said.
Griffin speculated that the IRS might one day develop its own generative AI tool to help analyze comment letters or generate private letter rulings, but it would have to be a closed system within the agency so that confidential taxpayer information doesn't become part of a publicly accessible database.
Unlike a static algorithm, AI learns and changes based on the information it receives, Griffin said, explaining that the agency would be very concerned with what AI would learn about its decision-making process.
"There would definitely be an interest on the part of the IRS in not exposing things like internal deliberations about guidance [or] some nonpublic data that it's using," she said. "So the possibility of an in-house system is intriguing in either the policy or the enforcement context."
The IRS is already using AI tools to inform the way it examines complex tax returns filed by wealthy corporations.
IRS Commissioner Daniel Werfel testified on Capitol Hill this week that the agency uses AI to strategize against tax evasion by multinational corporations that move the profit from their economic activity in the U.S. to foreign jurisdictions, thereby representing it as though it's not happening in the U.S.
Werfel said agency staffers are using a combination of in-house and private-sector AI tools to calculate various scenarios and advise employees on the best course of action against some of the largest corporations in the world that try to shield their income from taxes in very sophisticated ways.
"We're having the computer help us calculate all the permutations of the moves, so that we're better at the chessboard and can catch the evasion more effectively," Werfel said.
However, he emphasized that the final decision on what steps to take when auditing a large corporation remains with IRS employees.
Going forward, AI's role in federal tax regulations doesn't have to be limited to the federal government, wealthy corporations or those who can hire tax lawyers to make their case.
Minkovich said AI presents an opportunity to eliminate some of the bias in the code toward well-funded, sophisticated stakeholders such as businesses and trade groups by helping individual taxpayers participate in the regulatory process.
She said AI can help individual taxpayers understand technical jargon in tax regulations as well as the "somewhat specialized skill set" needed to navigate regulations.gov, the federal website that allows the public to comment on regulatory materials.
"You get more comments on regulations that impact businesses than you do on regulations that impact the individual side of the tax code," she said. "AI may remove some of the barriers to entry that John Q. Public may otherwise experience in trying to submit comments on tax regulations."
--Editing by Neil Cohen and Roy LeBlanc.
For a reprint of this article, please contact reprints@law360.com.