Interesting ruling! A New York court has ruled (at least orally) that AI created documents sent to counsel are not privileged!

The ruling by New York District Judge Jed S. Rakoff has been reported in several places, including on Law360 (subscription required) and by Michael Berman on the EDRM blog here. Hat tip to Tom O’Connor for giving me the heads up about it, sharing this LinkedIn exchange about the ruling.

The case involves a Texas financial services executive accused of a $150 million fraud. As the government motion for a ruling that the AI created documents sent to counsel are not privileged states, “the defendant had run queries related to the Government’s investigation through an AI tool (Claude) created by a third-party company, Anthropic. Defense counsel further informed the Government that documents generated by the AI tool reflecting Heppner’s prompts and the AI tool’s responses (that is, the AI Documents) would be located on the electronic devices that the Government had seized during the search. To date, defense counsel has identified approximately thirty-one documents in the Seized Materials which comprise the AI Documents. Counsel has asserted that such documents are privileged.”

The government’s argument against the application of privilege was structured into three primary legal challenges.

Failure to Meet Attorney-Client Privilege Standards

The government argued that the AI Documents failed every necessary element of attorney-client privilege as defined in the Second Circuit.

Element of Privilege	Government’s Rebuttal
Communication with Attorney	An AI tool is not a licensed attorney, holds no law degree, and owes no professional duties to the court or the client.
Purpose of Legal Advice	Anthropic’s terms and the tool’s own responses explicitly disclaim providing legal advice, instead directing users to consult qualified attorneys.
Confidentiality	The defendant voluntarily shared data with a third party whose privacy policy allows for data training and disclosure to “governmental regulatory authorities.”

Key Legal Precedent: The motion cited In re OpenAI, Inc., Copyright Infringement Litig. (2025), noting that “the discussion of legal issues between two non-attorneys is not protected by attorney-client privilege.”

The “Retroactive Cloaking” Argument

Defense counsel argued that because Heppner shared these AI Documents with them for the purpose of seeking legal advice, the documents became privileged. The government countered that:

• Preexisting, non-privileged materials do not become privileged simply by being sent to a lawyer.
• Transmitting a Google search or a library record to an attorney would not make the original search history privileged; the government likens the AI searches to these activities.
• While the advice given by the lawyer regarding the AI output may be privileged, the underlying AI Documents themselves remain discoverable.

Inapplicability of the Work Product Doctrine

The work product doctrine shields materials prepared “by or at the behest of counsel” in anticipation of litigation. The government noted:

• Defense counsel admitted they did not direct Heppner to run the Claude searches.
• The defendant acted on his own initiative as a layperson conducting independent research.
• The doctrine is intended to protect an attorney’s “zone of privacy” and thought processes; it does not protect a defendant’s independent use of commercial software.

Judge’s Ruling

During the hearing on Tuesday, Judge Rakoff reportedly stated: “I’m not seeing remotely any basis for any claim of attorney-client privilege.” And when Prosecutor Alexandra Rothman said the documents do not “reflect the legal strategy” of Heppner’s defense team, Judge Rakoff agreed, concluding that the AI documents were not prepared by attorneys. The judge also noted that the tool used by Heppner contains a provision that any information inputted is not confidential — essentially, that a user has no expectation of privacy.

However, the Judge reportedly “said that if prosecutors try to use the AI-generated information at trial, it could give rise to a ‘witness-advocate conflict’ since his law firm would become a witness in such a scenario.”

Observations

One observation about the confidentiality considerations: Citing and linking to Anthropic’s Privacy Policy, the government’s motion states: “Anthropic explicitly advises its users in its Privacy Policy, which was in effect at the time of the defendant’s searches, that it collects data on the ‘prompts’ entered and ‘outputs’ generated; that it uses this data to ‘train’ its AI tool; and that it may disclose this data to ‘governmental regulatory authorities’ and ‘third parties.’”

But regarding the training aspect, Anthropic’s policy states: “We will not use your Inputs or Outputs to train our models, unless:…(3) you’ve otherwise explicitly opted in to the use of your Inputs and Outputs for training purposes.” So, unlike OpenAI’s opt-out policy, Anthropic’s is opt-in. While we don’t know whether the defendant opted in, my guess is that it would be unlikely (in my opinion, most people would choose not to opt in). So, if Anthropic isn’t using the defendant’s data to train Claude, would that make a difference?

The expectation of privacy argument is also an interesting one. What if the defendant had simply typed up his thoughts in a series of Gmails and sent them to counsel? Would that have changed anything? Years ago, Google claimed in a filing that Gmail users have no reasonable expectation of privacy. Would Judge Rakoff have made the same ruling if these were Gmail created documents instead of AI created documents sent to counsel?

The government made enough arguments that had nothing to do with the fact that these were AI created documents sent to counsel that the ruling might have been the same if they weren’t. Regardless, it will be interesting to see if other courts find that AI created documents sent to counsel violates privilege, especially if other circumstances (such as counsel possibly directing their client to create them) are different.

So, what do you think? Should using AI to create documents waive privilege? Please share any comments you might have or if you’d like to know more about a particular topic.

Image created using Microsoft Designer, using the term “a soup server with dark hair and a dark mustache refusing to sell soup to a well dressed robot lawyer with a sign behind the server reading ‘No Privilege for You’”.

Disclaimer: The views represented herein are exclusively the views of the author, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.