In the past few months, a growing number of judges have started issuing orders that bar lawyers and parties from using public artificial intelligence tools during the discovery phase of litigation. These orders, which apply to widely used services like ChatGPT and similar chatbots, are not about technical incompetence or a fear of technology. They are about something more basic: privacy.

The reasoning is straightforward. When you paste a document or a set of facts into a free, public AI tool, you have no real assurance about what happens to that data afterward. It may be stored on servers in unknown jurisdictions, used to retrain the model, or accessed by third parties. In court proceedings, where confidentiality rules and protective orders govern sensitive information, that uncertainty is unacceptable.

What Happened – and Why

According to Bloomberg Law News reports, judges in several U.S. district courts have recently added explicit prohibitions against the use of “public-facing generative AI tools” in discovery orders. The orders typically cite the risk that confidential documents, personal data, or attorney work product could be exposed or become part of the AI model’s training data.

One judge noted that even deleting a conversation after using an AI tool does not guarantee the information is gone. The data may have been copied to backup systems or used to improve the service. Because these tools are not covered by standard legal confidentiality protections (like attorney–client privilege), the only safe approach is to keep them out of the process altogether.

The bans are not universal, and some courts have taken a softer approach—simply reminding parties to be cautious. But the trend is clear: the judiciary is waking up to the fact that free AI services operate under terms of service that give little assurance to users.

Why It Matters for Everyone, Not Just Lawyers

You might not be involved in a lawsuit, but the same risks apply to your daily use of AI tools. When you type a question about a health concern, paste a draft of a business proposal, or ask for help summarizing a personal email, you are sending that information to a company you likely know very little about.

Most public AI services are financed through user data, either directly by selling access to the information or indirectly by using it to improve the product. Privacy policies are often vague about retention periods, encryption standards, and third-party sharing. A 2024 analysis by the Mozilla Foundation found that several popular generative AI tools could be described as “privacy nightmares” because of how they handle personal data.

The key point is that there is no enforceable promise of confidentiality when you use a free, public AI tool. Even if the company states it will not share your data, nothing prevents an employee or a contractor from viewing it. In the United States, there are currently no federal laws that comprehensively regulate how AI companies handle user data.

What You Can Do to Protect Your Privacy

You do not have to stop using AI entirely. But you can make smarter choices about what you share and which tools you trust.

Avoid sharing personally identifiable information. Do not paste in your full name, address, phone number, Social Security number, or financial account details. Treat any conversation with a public AI tool as something that could be read by a stranger.

Do not paste confidential documents. If you would not send a document to a random email address, do not paste it into a chatbot. For legal professionals, the rule should be even stricter: unless you are using an enterprise version with a data processing agreement that guarantees confidentiality, do not upload anything covered by a protective order or privilege.

Consider using local AI models. You can run open-source large language models on your own computer using software like Ollama or LM Studio. These models are not as powerful as the latest commercial systems, but they keep all your data offline. For day-to-day tasks like drafting text or summarizing notes, they are often sufficient.

Check the privacy policy. Before using any AI service, look for clear language about data retention, encryption, and whether the company uses your inputs for training. If the policy is vague or says “we may share data with third parties,” assume the worst.

Use services that offer contractual protections. Some AI providers sell enterprise versions that include data processing agreements, encryption at rest and in transit, and guarantees that your data will not be used to train the model. These are the only safe options for handling sensitive or confidential information.

Sources

The information about judicial bans on AI tools during discovery comes from Bloomberg Law News (June 2026). Their reporting is behind a paywall, but the key points have been summarized in legal blogs and professional newsletters. For a broader view of AI privacy risks, the Mozilla Foundation’s “Privacy Not Included” guide to AI chatbots offers a consumer-friendly look at major services.

The bottom line is this: courts are ahead of the general public in recognizing that free AI tools are a privacy risk. Their reaction may seem extreme, but it is based on real gaps in how these services handle data. Whether you are a lawyer, a student, or just someone who uses ChatGPT for entertainment, it is worth applying the same caution to your own information.