A new publication from The Sedona Conference Journal says: “The Machine Isn’t the Interlocutor”. What does that mean? Let’s find out.
Yesterday, The Sedona Conference Journal announced a critical analysis of the recent decisions in U.S. v. Heppner and Warner v. Gilbarco, two very different opinions on whether a party’s use of GenAI tools forecloses any claims of attorney-client communication privilege or work product protection.
The publication is titled The Machine Isn’t the Interlocutor: Why United States v. Heppner Gets Privilege Wrong, and it’s available here. The article is co-authored by Bridget McCormack, who is the President and CEO of the American Arbitration Association and the former Chief Justice of the Michigan Supreme Court and Shlomo Klapper, the founder and the CEO of Learned Hand, an AI platform designed for judges and courts.
As the title states, they liked one decision more than the other. Among their issues they had with the Heppner case:
The Anthropomorphization of AI
The court treated Claude as an “interlocutor”: a third party with independent agency. In reality, Claude is a large language model that predicts statistically likely tokens.
- Lack of Agency: Unlike a human third party, AI cannot be deposed, cannot decide to contact the government, and has no volition to betray a confidence.
- Technical Precedent: AI should be viewed similarly to calculators or search engines—sophisticated tools that respond to inputs but do not constitute “third parties” in a legal sense.
Misapplication of Waiver vs. Independent Privilege
The Heppner court asked whether AI interactions independently satisfy the elements of privilege. However, the correct inquiry is whether interacting with a tool about already privileged information constitutes a waiver.
- The “Alchemy” Counter-Argument: While non-privileged documents do not become privileged by being shared with counsel, the reverse is also true: privileged material is not “alchemically” stripped of its status simply by being processed through a computational tool.
Conflict with Cloud Computing Norms
The court’s focus on Anthropic’s Terms of Service (ToS) as a breaker of confidentiality contradicts existing ethics and legal standards:
- ABA Formal Opinion 477R: Established that lawyers may use cloud services (which have a similar ToS regarding data access) without violating confidentiality, provided reasonable precautions are taken.
- Stengart v. Loving Care Agency: Confirmed that a user’s reasonable expectation of confidentiality is not defeated by a service provider’s ToS alone.
- Functional Parity: Under Heppner’s logic, any document stored in Google Drive or any email sent via Outlook would lose privilege because the providers technically reserve rights to process and disclose data.
Broader Legal Consequences
The Heppner precedent creates several systemic risks:
- Disruption of Necessary Intermediaries: The court’s logic could be extended to human intermediaries (like translators) if their engagement was not specifically directed by counsel.
- Harm to the Least Resourced Litigants: Litigants who use AI to process their lawyer’s advice into draft declarations or strategies would find their entire preparation history discoverable by adversaries.
- Prosecutorial Asymmetry: While federal prosecutors use AI for investigation and case preparation under the protection of deliberative-process and work product privileges, defendants using the same technology would be forced to create a “timestamped, verbatim record” of their preparation for the prosecution.
The Proposed Functional Test
Instead of asking if a tool is a “third party,” courts should ask if the technology interaction creates specific risks that the third-party disclosure rule was designed to address:
- Testimony: Is the entity capable of being deposed or testifying?
- Compelled Production: Can the entity be served with a subpoena as a legal person?
- Volitional Disclosure: Does the entity have independent legal standing and the agency to disclose information of its own accord?
The authors conclude this:
“Heppner answers a new application of an old question. But its misstep at the outset in anthropomorphizing a computational tool causes it to see a third-party disclosure where there is none. Because it sees a communication, it frames the case as an inquiry into whether the AI interactions independently satisfy the elements of privilege. Using computational infrastructure has never been characterized as a communication, and therefore never waived the attorney-client privilege. Large language models can do more than previous technologies, but they are still technologies. Privilege in the inputs carries into the outputs.”
The 20-page article is available here.
So, what do you think? Do you agree that the machine isn’t the interlocutor? Please share any comments you might have or if you’d like to know more about a particular topic.
Image created using DALL-E 3, using the term “a soup server with dark hair and a dark mustache selling soup to a well dressed robot lawyer with a sign behind the server reading ‘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.