Law360, London (March 16, 2026, 8:03 PM GMT) -- The use of artificial intelligence tools is expected to spark novel battles for disclosure during litigation, with opposing parties likely to clash over the confidentiality of AI systems and whether their use has undermined privilege.

Recent court decisions in the U.S. and Britain suggest that litigators will increasingly need to grapple with questions over the disclosure of documents produced with the help of AI. Lawyers have also suggested that major tests for privilege could become flash points as litigators get to grips with the new technology.

Ian De Freitas of Farrer & Co. LLP said he believed it will soon become "pretty standard" for litigators to question their adversaries' use of AI and mount arguments over whether material prepared with AI can be disclosed.

"Litigants will probe whether there has been information surfaced in the context of the dispute which is not privileged, so because it's not between a lawyer and a client, or it doesn't fall within that litigation privilege wrapper," he said. "You're going to get this coming up as a pretty standard question as we move forward."

De Freitas and others think that a major battleground will be confidentiality, which underpins privilege.

Litigators are likely to ask during the disclosure stage of litigation, "Have the lawyers, or has your client used AI in any way in relation to this case? And please disclose how they've used it and what system they used, and crucially, was it confidential?" he said.

A judge ruled on Feb. 10 at a New York federal court that an exchange between a defendant in a fraud case and Anthropic's Claude AI chatbot was not protected by privilege.

A Manhattan federal court held that financial services executive Bradley Heppner could not claim privilege over a defense strategy he drafted with Claude because, among other things, Claude's privacy policy stated the tool was not confidential. Heppner is facing allegations of defrauding investors out of more than $150 million. 

There has not yet been an English case dealing with the same issues. But lawyers have said the case is illustrative of the kinds of arguments that are likely to be run in the U.K.

Tom Whittaker, head of AI advisory at Burges Salmon LLP, said the decision showed there was "potentially a strong argument" that use of certain public AI systems will mean there is no confidentiality, and therefore no privilege.

"It's an example of where, A, that argument has actually been run, and B, the court has found for it, but it also shows the specific way that it was run, which is by looking specifically at the terms of use for that AI system," he said.

A judgment handed down in February by the Upper Tribunal also highlighted the perils of using public AI systems in a case that involved two immigration lawyers who had submitted filings with what may have been AI-hallucinated citations. 

The tribunal in that case did not rule on issues of confidentiality or privilege, but a panel of three judges gave a stark warning about putting confidential information into AI chatbots.

"Uploading confidential documents into an open-source AI tool, such as ChatGPT, is to place this information on the internet in the public domain, and thus to breach client confidentiality and waive legal privilege," the judgment said.

De Freitas said that an AI tool's terms of use wouldn't necessarily be the end of a dispute about confidentiality.

Parties to litigation could get drawn into disputes about whether AI systems are truly confidential and whether their confidentiality was compromised by data escaping to third parties, a process De Freitas described as "leakage."

"These AI tools say that they're closed, but are they?" he said. "I do worry sometimes that some of these systems, when they say that they're closed, nevertheless, information could leak and if you ask the right sort of questions, you might find out that actually it's not as closed as people think it was."

Whittaker said that pro-privilege arguments would be strengthened where the AI systems used have "both contractual and technical protections around confidentiality, privacy and security."

The New York federal court also ruled that Heppner's interactions with Claude were not privileged because they weren't covered by the work product doctrine, which provides some protection for material prepared by or for counsel in anticipation of litigation.

The court ruled that even if the material generated by Claude was prepared "in anticipation of litigation," it was not prepared by or at the request of Heppner's counsel.

There is no analogous concept to work product doctrine in English law, but arguments could instead revolve around the tests for litigation privilege, which applies to confidential documents prepared for the dominant purpose of litigation and when litigation is ongoing or reasonably anticipated.

De Freitas said that disputes over whether materials were generated for the dominant purpose of litigation could arise in cases where litigation has followed an internal investigation by a company into wrongdoing.

Those battles are likely to be fought over where the dividing line lies between AI use for conducting the investigation and building a case for litigation, he said.

"You could be using the AI, first of all, to give you a broad overview of what's happened — that's not privileged because the dominant purpose is to stop the thing from happening again," De Freitas said. "But then the second use of the AI to do a forensic review with a view to pursuing bad actors, and that may well be with a view to litigation against them, and therefore that is privileged."

Whittaker said that a major difficulty for a party seeking to challenge privilege on those grounds will be finding evidence to undermine a party's claim that documents were prepared for the dominant purpose of litigation. That is because, unlike the questions of confidentiality, where terms and conditions could be useful and easily available, the dominant purpose would be harder to determine without seeing the documents themselves, he said.

"They'll have to point to other evidence or factors surrounding it to try and make the claim that it's arguable that there was no dominant purpose in those circumstances," he said. "It may be necessary then to take additional steps with the court, such as asking the court to then look at that material in a protected way to be able to determine whether there was that dominant purpose."

To complicate matters further, the contents of a single document might not be enough to build up an evidential picture to challenge privilege, Whittaker said.

Professional bodies and the judiciary have sought to issue guidance on the use of AI, with the Bar Council updating its guidance in November. The judiciary published updated guidance for the use of AI by judges and tribunal members in October. 

Whittaker said clients should speak with their lawyers to understand what steps to take to maintain privilege and take a moment to think before using an AI tool.

"If you are going to use AI," he said, "it needs to be in the right circumstances for the right purposes with the right tool — one where you are confident that there are protections for confidentiality so that you understand what happens with the data and so that you understand what you will do with that data or the output afterward."

--Editing by Robert Rudinger.

For a reprint of this article, please contact reprints@law360.com.