In a February opinion, U.S. District Judge Jed Rakoff ruled that former GWG Holdings Chairman Bradley Heppner had waived confidentiality protections under attorney-client privilege for information he had shared with an AI chatbot without his counsel's consent. Heppner was convicted May 7 of securities fraud and other charges for using shell companies to extract $150 million from GWG.
A New York federal court ruling that denied privilege to a client's communications with an AI platform is among the first to address attorney-client privilege protections for legal discussions with such platforms. (iStock.com/Dmitry Nogaev)
Although the opinion arose from a securities fraud prosecution, its reasoning on AI communications could extend broadly to the dynamics between tax advisers and their clients — a situation that underscores the need for firms and companies with tax departments to reevaluate their technology policies and client agreements to preserve privilege protections, experts told Law360.
"As attorneys, we now need to address our AI policy in our engagement agreements and other procedural rules," said Travis Thompson of Fennemore Craig PC.
In just a few years, rapid advancements in AI have expanded into the legal industry, touted as improving workflows at firms and companies while potentially broadening access to legal services for people who lack resources.
In the tax practice, the technology has been used with the aims of automating routine compliance tasks, streamlining document processing and accelerating research and analysis of tax code sections.
The Internal Revenue Service has also expanded its use of AI as the agency reduces its workforce, using the technology to identify potentially risky tax returns for audit, detect fraud and communicate with taxpayers through chatbots.
And as AI use has gained popularity, companies across industries have developed their own AI policies and conveyed them to clients and partners. For example, some law firms allow the use only of certain AI tools in their workflows.
Thompson said firms could go a step further. For example, he said his firm is updating its AI policy letters, citing the Heppner decision and its implications on certain discussions between clients and attorneys.
At this point, attorneys should inform their clients about their waiver of attorney-client privilege if they use AI platforms to research their case without the direction of the attorney or face a potential malpractice suit, he said.
That AI tools are not foolproof has been highlighted by courts in recent instances of potential misuse of the technology.
The U.S. Tax Court, for example, has encountered factual errors in briefs and reports that judges said likely came from unverified information generated by AI platforms.
Although the Tax Court has yet to impose stiff sanctions for misusing AI, practitioners said the public embarrassment, as well as the reputational harm, attorneys and other experts face for failing to catch the errors is punishment enough.
Nevertheless, the Heppner decision adds to growing concerns that AI use may outpace the legal safeguards designed to protect sensitive tax and legal communications, as Judge Rakoff warned in his opinion.
"AI's novelty does not mean that its use is not subject to longstanding legal principles, such as those governing the attorney-client privilege and the work product doctrine," the judge said.
In the tax practice, privilege is paramount to certain tax advice. That is why eligible accountants, enrolled agents and other professionals authorized to practice before the IRS are also obligated to protect such confidential information under Internal Revenue Code Section 7525
.Although the Heppner decision did not address tax advice privilege, and no tax cases have yet examined confidentiality issues involving AI communications, the ruling underscored the need to consult attorneys or other authorized practitioners for tax and legal advice, according to Conor Desmond of Caplin & Drysdale.
"Keeping a human in the loop does allow for more protection under privilege" than getting an answer from an AI platform, he said.
In the New York case, Heppner asserted privilege over documents containing information addressing last year's grand jury investigation into him that he had shared with an AI model. He argued the documents were confidential because they contained information about legal strategy he discussed with his attorney.
Heppner used Claude, a popular large language model developed by Anthropic, which was founded in 2021.
In disagreeing with Heppner, Judge Rakoff ruled that one of the reasons the AI communications could not be confidential is because Claude is not a human with fiduciary duties — and privilege requires a trusted relationship with a human.
"No such relationship exists, or could exist, between an AI user and a platform such as Claude," the judge said.
Furthermore, Heppner's attorney did not direct his client to use Claude on matters involving the grand jury investigation, which the judge said meant the information Heppner shared with the platform neither constituted legal advice nor reflected the attorney's legal strategy protected by the work product doctrine.
Experts said the Heppner opinion also shows how tax practitioners can emphasize their value amid the rise of AI technology.
A key takeaway from the decision is that the technology should support tax advisers, not replace them, according to Aaron Esman of Ziering & Esman PLLC.
"This is a tool, not a replacement," he said.
Because a vast majority of sophisticated clients already use AI tools, attorneys increasingly must distinguish themselves through their expertise and experience, according to Del Wright, a tax professor at Louisiana State University's Paul M. Hebert Law Center.
"This is the new game, and so the question is, 'How do we structure this to best benefit the client?'" Wright said.
Still, law and accounting firms need to be better at advertising this to their clients and incorporating their AI policy into agreements, according to Thompson of Fennemore Craig.
"Many firms have not addressed AI in their engagement letters, and they absolutely 100% need to," he said.
--Editing by Aaron Pelc and Lief Nielsen.
For a reprint of this article, please contact reprints@law360.com.
