Why Some Judges Are Banning Public AI Tools in Court – and What It Means for Your Privacy

Why Some Judges Are Banning Public AI Tools in Court – and What It Means for Your Privacy

In recent months, a quiet but significant trend has emerged in U.S. courtrooms: judges are formally barring the use of public generative AI tools like ChatGPT and Google Gemini during the discovery phase of litigation. The stated reason is privacy. These rulings, reported by Bloomberg Law in early June 2026, reflect a growing judicial awareness that public AI platforms pose real risks to the confidentiality of sensitive information—risks that extend far beyond the legal profession.

If you use any of these tools for work, personal projects, or even casual conversation, the same vulnerabilities apply to your data. Here’s what the court orders reveal, why they matter, and what you can do to protect yourself.

What Happened

Multiple federal and state judges have issued orders prohibiting attorneys and parties from inputting discovery materials—documents, emails, deposition transcripts, and other confidential records—into public AI services. The concern is not about the AI’s accuracy but about what happens to the data after it is submitted.

According to privacy experts cited in the Bloomberg Law article, these services often retain user prompts, and some use them to train or improve their models. This means that anything entered into a public AI chat could be stored, reviewed by employees or contractors, or even used to generate responses for other users. For legal cases, that is a direct threat to attorney-client privilege and to protective orders that govern how sensitive information is handled.

The judges’ bans are not a blanket prohibition on all AI use in litigation; they target specifically the public, free or low-cost versions of tools that do not offer enterprise-level privacy guarantees. Some orders explicitly allow the use of locally installed models or secure, contracted AI services that sign data processing agreements with law firms.

Why It Matters for Everyone

The logic behind these judicial orders applies just as strongly to your own use of ChatGPT, Gemini, or similar products. When you paste a draft of a personal email, ask for advice on a work project that involves proprietary numbers, or describe a health concern in detail, that information leaves your control.

Most public AI services’ privacy policies explain that they may collect, retain, and use input data for training. Terms of service often include clauses that grant the company broad rights to use the content you provide. While many providers now offer opt-out options for training data, those settings are not always obvious or turned on by default.

A Bloomberg Law article from June 5, 2026, notes that the current ecosystem of public generative AI tools lacks the kind of confidentiality guarantees that professionals—and increasingly, judges—expect. If a court considers these tools too risky for discovery materials, the average user should think twice before entering anything they would not want seen by a stranger or stored indefinitely on a company’s servers.

There is also a lack of transparency about data retention periods. Even if you delete a conversation, the service may keep logs for compliance, security, or model improvement purposes. Exactly what is retained, for how long, and by whom is often unclear.

What You Can Do to Protect Your Privacy

You do not have to stop using generative AI tools entirely, but you can take practical steps to limit your exposure.

Use local or offline models when possible. Several open-source language models can run entirely on your own computer or phone, sending no data to external servers. Tools like Ollama or LM Studio allow you to download and run models locally. The trade-off is lower performance on weaker hardware, but the privacy gain is significant.

Check the privacy settings of the tool you are using. For ChatGPT, you can access the “Settings & Privacy” menu and find the option to “Improve the model for everyone.” Turning this off prevents your conversations from being used for training, though the company may still retain data for other purposes. On Google Gemini, you can similarly turn off “Activity & personalization.” However, these settings change over time, so re-checking periodically is wise.

Never paste sensitive information into a public prompt. Treat any text you enter as if it might be published online. That includes financial details, personal identification numbers, medical history, private correspondence, and trade secrets. Even seemingly harmless information can be pieced together by an adversary or by the model itself in future interactions.

Review the terms of service and privacy policy for each tool you adopt. This is tedious but important. Many users click “agree” without reading. Know what rights you are granting the provider. If the policy is vague or grants broad data use permissions, consider whether the convenience is worth the risk.

If you are a professional handling confidential data (legal, medical, journalism, etc.), assume public AI is not allowed unless your organization has vetted a secure, contractually protected option. The emerging court guidance is a clear signal: the default presumption should be against submitting private material to these tools.

Future Outlook

The judicial orders are likely just the beginning. As generative AI becomes more embedded in daily work, we can expect more formal guidance from courts, professional ethics boards, and regulators. In the meantime, the responsibility for data privacy falls mainly on the user. The same risks that worry judges apply to anyone who types personal or confidential information into an AI interface.

Staying informed about how each platform handles your data, and choosing tools that respect your privacy, is the most practical defense. The courts have drawn a bright line for discovery. For the rest of us, the boundary may be less visible, but the underlying caution is the same.

Sources

• Bloomberg Law News, “Judges’ Public AI Bans During Discovery Zero in on Privacy Risk” (June 5, 2026). Paywalled article, summarized via Google News RSS.

Note: Specific judicial orders and exact language from the rulings were not independently verified beyond the Bloomberg Law summary. Privacy policies vary by provider and change over time; readers should always consult the latest version directly.