The Machine Isn’t the Interlocutor: EDiscovery Trends

Key Takeaways

• Heppner treats AI as a third‑party, contrary to established privilege doctrine
• Privilege should persist when AI merely processes already privileged information
• Court’s reliance on AI Terms of Service threatens cloud‑service confidentiality
• Proposed functional test evaluates deponibility, subpoena power, and volitional disclosure

Pulse Analysis

The rapid integration of generative AI into eDiscovery has forced courts to confront old privilege rules in new settings. In U.S. v. Heppner, the district court labeled Anthropic's Claude as an "interlocutor," effectively treating the model as a third party capable of waiving attorney‑client privilege. The Sedona Conference Journal’s response, authored by former Michigan Supreme Court chief justice Bridget McCormack and AI‑platform founder Shlomo Klapper, argues that this anthropomorphization misapplies the law. By likening AI to calculators or search engines—tools that lack agency and cannot be deposed—the authors assert that privilege should remain intact when privileged material is simply processed by a computational system.

Beyond the doctrinal debate, the Heppner ruling collides with established ethical guidance such as ABA Formal Opinion 477R, which permits lawyers to use cloud services provided reasonable safeguards are in place. The court’s focus on Anthropic’s Terms of Service risks extending privilege loss to ubiquitous platforms like Google Drive or Outlook, undermining confidentiality expectations across the legal industry. The authors highlight that this could create a chilling effect for litigants who rely on AI to draft filings or strategize, especially those without deep resources, while prosecutors continue to leverage the same technology under existing work‑product protections.

To resolve the tension, the Sedona paper proposes a functional test that asks whether the technology can be deposed, served with a subpoena, or independently disclose information. This pragmatic framework shifts the analysis from abstract agency to concrete legal risks, preserving privilege while acknowledging AI’s role as a sophisticated tool. As courts grapple with AI’s expanding footprint, adopting such a test could provide clearer guidance, protect client confidentiality, and maintain a level playing field for all parties in the evolving eDiscovery landscape.