Judges Are Banning Public AI in Discovery – Here’s Why Privacy Is the Core Concern

In recent weeks, a growing number of federal and state judges have issued orders restricting the use of publicly available AI tools—such as ChatGPT, Claude, or Gemini—during discovery in civil litigation. These orders, reported by Bloomberg Law on June 5, 2026, are not blanket prohibitions of all AI, but they zero in on the privacy risks that arise when sensitive legal documents are processed through cloud-based models.

If you are a lawyer, paralegal, or litigant handling discovery, these rulings are worth paying attention to. They signal that courts are now looking beyond convenience and cost savings to ask a harder question: what happens to your data after you paste it into a public AI tool?

What Happened

Judges in at least three separate cases—two in federal district courts and one in a state superior court—have recently ordered parties not to upload discovery materials to public generative AI services without prior approval or without implementing specific confidentiality safeguards. The reasoning is consistent: once data enters a public AI system, it may be used to train models, stored on servers outside the court’s jurisdiction, or exposed to third parties. Even if the AI provider claims not to retain data, the risk of inadvertent disclosure remains higher than with traditional discovery tools.

For example, one judge noted that discovery often contains trade secrets, medical records, and personally identifiable information. Uploading such material to a free-tier AI service could violate protective orders and ethical duties of confidentiality. Another judge pointed out that litigants might unknowingly share privileged communications.

A related Bloomberg Law article from January 2026 also flagged pitfalls with AI notetaking in depositions, where recordings and transcripts fed into cloud-based AI tools raise wiretapping and discovery concerns. The new discovery orders build on that same logic: the convenience of AI should not override the core duty to protect client information.

Why It Matters

The privacy risks behind these bans are not hypothetical. Public AI tools typically store user input to improve their models or at least process it through external servers. For a law firm, uploading a contract with proprietary pricing or a medical record with a client’s Social Security number could violate state data breach laws, ethical rules under ABA Model Rule 1.6, or even HIPAA. Moreover, if opposing counsel can later subpoena the AI provider’s logs, the entire discovery process could be undermined.

For litigants, the implications are practical: if your lawyer uses a public AI tool to summarize documents or draft discovery requests, you may unknowingly lose control over sensitive information. Courts are essentially saying that the default settings of these tools are not compatible with the confidentiality requirements of litigation.

What Readers Can Do

If you are involved in a case where a judge has already banned or restricted public AI use, the first step is to read the order carefully. Some allow AI tools if they are enterprise-grade with contractual data protection clauses. Others require a specific stipulation from both parties.

For legal professionals, consider these practical measures:

  • Avoid using free-tier AI tools for any discovery-related work. Assume that anything you paste into a public chat interface could be visible to the provider or used for training.
  • Use secure, litigation-specific AI platforms that offer data isolation, encryption, and audit trails. Many e-discovery vendors now include AI features that are designed to comply with confidentiality rules.
  • Draft a standing order or protocol with opposing counsel about acceptable AI use in discovery. Addressing this early can prevent a dispute later.
  • Review your ethical obligations. ABA Formal Opinion 512 (2024) and similar state opinions hold that lawyers must take reasonable steps to prevent disclosure of confidential information when using technology. Relying on a public AI tool without verifying its data practices may not satisfy that standard.

For privacy-conscious consumers involved in litigation, ask your lawyer what tools they are using for document review and whether those tools store or share your data. You have a right to know.

Broader Implications

These court orders are likely just the beginning. As generative AI becomes cheaper and more powerful, judges will continue to weigh efficiency against privacy. We may see model discovery orders that explicitly bar the use of public AI, or create a checklist of acceptable safeguards. Some bar associations are already drafting guidance.

The key takeaway is that public AI and legal confidentiality do not mix by default. Until the technology and regulations catch up, the safest path is to treat all public AI tools as hostile territory for your client’s data.

Sources

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