Why Judges Are Banning AI in Court Discovery – And How to Protect Your Privacy

Introduction

The use of generative AI tools like ChatGPT, Claude, or Microsoft Copilot has become common in legal workflows—especially during discovery, where large volumes of documents need to be reviewed, summarized, or categorized. But a growing number of judges are now stepping in to restrict or outright ban the use of public AI tools during discovery. The reason? Privacy risks. Recent rulings highlight that uploading confidential case materials into consumer‑grade AI systems can expose sensitive data to third‑party servers, create unintended wiretapping liabilities, and raise ethical concerns about attorney‑client privilege.

If you’re a lawyer, paralegal, or simply someone who uses AI for note‑taking or document handling in a sensitive context, understanding what’s happening in courtrooms will help you avoid costly mistakes—and protect your own privacy.

What happened

In early June 2026, Bloomberg Law News reported that judges in several U.S. jurisdictions have issued orders prohibiting parties from using public generative AI tools during discovery. The orders typically cite a core worry: when you paste a deposition transcript, contract, or medical record into a free online AI tool, that data may be transmitted to servers outside the legal team’s control. It may be stored, used for model training, or potentially exposed in a data breach.

For example, one federal magistrate judge in a trade secrets case required all parties to “certify that no generative AI tool that operates on external servers has been used to review or summarize any discovery materials.” Similar orders have appeared in state courts as well. The judges are not banning all use of AI—only the use of tools that send data to public cloud services without adequate privacy protections.

At the same time, a separate Bloomberg Law article from January 2026 warned that AI notetaking tools (like transcription add‑ons on calls or meetings) can create legal problems beyond discovery. In some states, recording a conversation without consent is illegal, and an AI tool that automatically transcribes and stores conversations may count as a “wiretap.” Such tools also produce records that can be discovered in litigation, which might expose sensitive discussions a user never intended to preserve.

Why it matters

The rulings and ethical alerts underscore a fundamental tension: generative AI is powerful for handling large amounts of text, but legal and personal privacy protections haven’t caught up. For lawyers, uploading client data to a public AI tool can waive attorney‑client privilege if a court later determines the data was disclosed to a third party. Even if no technical waiver occurs, the risk of a data breach or a subpoena targeting the AI provider’s servers is real.

For everyday users—small business owners, journalists, or anyone handling sensitive personal information—the same risk applies. Using an AI note‑taker during a confidential phone call or pasting a private financial document into a chatbot may seem harmless, but you are effectively sharing that information with a company whose data‑handling practices you may not fully understand.

Beyond privacy, there is also the question of accuracy. AI tools can hallucinate or misrepresent content. In a legal setting, relying on an inaccurate summary of a document could lead to flawed arguments or missed evidence. Judges are increasingly aware that AI‑generated work product must be verified, and that verification is harder when the AI operates as a black box.

What readers can do

If you work in a legal or other sensitive field, here are practical steps to reduce risk:

  • Use dedicated enterprise‑grade AI tools that offer data isolation, no third‑party model training, and contractual privacy guarantees. Many cloud providers now have “private” instances of their AI services (e.g., AWS Bedrock private endpoints, Microsoft Azure OpenAI with data residency). Check whether your firm or organization already has a secure AI platform.
  • Never paste raw confidential data into free chatbots. Even if the tool says it does not store data, policies can change, and a breach is not impossible. When you must use generative AI, redact names, case numbers, and other identifiers first.
  • Disable transcription and AI note‑taking on phone and video calls unless you have explicit consent from all participants and understand where those recordings are stored. Consider using a local recording tool that never touches the internet.
  • Establish a clear internal policy for AI use during discovery. Include guidelines for what types of documents can be uploaded to which tools, and require team members to log any AI tool they use for case work.
  • Stay informed about local rules. Some court orders explicitly list prohibited services. Ask your judge or opposing counsel early in the case whether any AI restrictions apply.

For everyday users: treat any public AI tool like a public Wi‑Fi network. Don’t share anything you wouldn’t want a stranger to read. Use end‑to‑end encrypted services when handling truly private information, and remember that even “temporary” chats can be stored server‑side.

Sources

  • Bloomberg Law News, “Judges’ Public AI Bans During Discovery Zero in on Privacy Risk” (June 5, 2026).
  • Bloomberg Law News, “AI Notetaking Poses Wiretapping, Discovery, and Ethical Pitfalls” (January 28, 2026).