Bradley Heppner likely could have afforded to retain any lawyer of his choosing, having allegedly defrauded investors out of $150 million. But he chose a somewhat different path. Whether it was because he believed himself to be better, or more motivated, than a lawyer, or was just a hands-on kinda guy, who knows. but he took to Anthropic’s AI bot, Claude, to research and develop his defense, producing about 31 documents on the subject.
When the government executed a search warrant, guess who ended up with the 31 docs? SDNY Judge Jed Rakoff held that they were not privileged.
It is well established that the attorney-client privilege attaches to, and protects from disclosure, “communications (1) between a client and his or her attorney (2) that are intended to be, and in fact were, kept confidential (3) for the purpose of obtaining or providing legal advice.” Courts construe the attorney-client privilege narrowly because it operates as an exception to the rule that “all relevant proof is essential” for a complete record and for “confidence in the fair administration of justice.”
Applying these principles here, the AI Documents lack at least two, if not all three, elements of the attorney-client privilege.
The first two elements are fairly obvious. Claude isn’t a lawyer. Claude isn’t a person. When a defendant “chats” with an AI bot, he is not talking with counsel (some of you might want to consider the implications for your sex life as well, and not with lawyers, please), but with a computer.
The second element is abrogated as well by the fact that the communications are not confidential, but available to third parties, from the AI company to anybody the AI company wishes to share them with, such as the government. Just as you should never put anything in an email that you wouldn’t tell a federal agent, don’t discuss it with Claude, either.
But the third element presents a more interesting question.
Third, Heppner did not communicate with Claude for the purpose of obtaining legal advice. This issue perhaps presents a closer call because Heppner’s counsel asserts that Heppner communicated with Claude for the “express purpose of talking to counsel.” But, as Heppner’s counsel also conceded, Heppner did not do so at the suggestion or direction of counsel.
Had counsel directed Heppner to use Claude, Claude might arguably be said to have functioned in a manner akin to a highly trained professional who may act as a lawyer’s agent within the protection of the attorney-client privilege. But because Heppner communicated with Claude of his own volition, what matters for the attorney-client privilege is whether Heppner intended to obtain legal advice from Claude, not whether he later shared Claude’s outputs with counsel. And Claude disclaims providing legal advice. Indeed, when the Government asked Claude whether it could give legal advice, it responded that “I’m not a lawyer and can’t provide formal legal advice or recommendations” and went on to recommend that a user “should consult with a qualified attorney who can properly assess your specific circumstances.”
Clearly, AI is incapable of providing legal advice no matter what the tech titans claim as it eviscerates low level associates who can’t draft a memo on their own. But, just as a lawyer can engage the services of a paralegal or investigator to fulfill aspects of the legal function as an agent of the lawyer, and thus covered by the umbrella of the privilege, why not Claude too?
The AI disclaimer, like most disclaimers, provides no more useful information than do the terms of service that no one reads, since Claude is obviously not a lawyer but purports to fulfill many of the functions of a lawyer. Pull up a decision on Google these days and a popup window immediately queries whether you want generative AI to provide a summary. Can the bot be trusted to provide an accurate summary? Beats me, as I’ve never pushed the button, but it would come as no surprise that a non-lawyer will prefer a summary to reading a decision.
With the internet simultaneously asking and qualifying its utility as a cheap source of legal information, both reliable and not to be relied on simultaneously, what’s a non-lawyer to do? Perhaps the better answer, per Judge Rakoff’s ruling, is that a pop-up window appears informing the user that any queries to Claude are confessions that can and will be used against him in court. But then, that wouldn’t be very good for tech business model.
Discover more from Simple Justice
Subscribe to get the latest posts sent to your email.
Edward Burns.
“There has been a great proliferation of lawyers in the past 20 years, just as there has been a proliferation of computers. But unlike computers, lawyers do not get twice as intelligent and half as expensive every two years.”
It’s a machine built to tell you what you want to hear. That is fundamentally opposed to what a lawyer does. Clients are wrong all the damn time about everything. This guy too, and he’s getting what his hubris deserves.
It’s pernicious for anyone thinking they are intelligent. Humility is required because the godforsaken things are going to end up agreeing with you and whispering lies, such as “You are so so smart; it’s everyone else who is wrong.” We aren’t built to withstand that forever especially when desperate enough.
“A man’s got to know his limitations,” is truer now than ever. I’d go further. Admit them, embrace them, absorb them. And please for the love of anything think critically.
Heck, *I* would prefer a summary to some decisions I’ve had to read.
Though, to be fair, the Oregon appellate system tends toward a direct, clear, and concise writing style.
What’s a non-lawyer to to? Recognize the old adage – “You get what you pay for.”
Like our host, I’ve never “pushed the button.” For that matter, given the commentary I’ve seen in various blawgs regarding AI in the legal system, I don’t see myself ever DOING a button-push.
But I’m a dinosaur like that
I agree that this wouldn’t be attorney-client privilege material, but it seems like it should have been a much closer question on whether it is protected by the work product doctrine than the court treated it as? I mean, I’ve never had occasion to research the issue in any jurisdiction, but I’d have to presume that, generally speaking, a pro se litigant can claim work product protection over their own research materials, witness outlines, etc. And yes, apparently Heppner actually had a lawyer at the time, but it seems really to undercut the idea of a defendant “being able to assist in his own defense” if that means that a represented defendant can ONLY safely conduct research and draft notes on legal strategy if it’s being done at the direction of counsel since that prevents the client from doublechecking the lawyer’s work or otherwise effectively supervising them.
Even if the output is arguably work product, Claude’s output is not confidential and therefore any A-C or work product privilege was waived.
A possible workaround would be for the company to sell a standalone version of Claude (“MAUDE?”) to be installed on the user’s computer under a contract where the user (client) assumes exclusive liability for any damages or injuries resulting from Maude’s use, where the seller relinquishes all rights to access any output from Maude and where prior to purchasing the software the user (client) has a one hour “consultation” with a $25/hr attorney retained by seller who “directs” client to explore their legal problem further with the AI which user then purchases on advice of counsel.