Judges Are Banning AI in Court Cases – Here’s What It Means for Your Privacy
If you’ve used ChatGPT, Google Gemini, or any other public AI tool to help draft a document, summarize a report, or even just brainstorm ideas, you might assume your conversations are private. A growing number of federal judges are challenging that assumption—and their reasons have implications far beyond the courtroom.
Recent court orders are specifically barring lawyers from using public generative AI tools during the discovery phase of litigation. The stated reason: the privacy risk is too high. And the same risk applies to anyone who feeds sensitive personal or business information into these tools.
What happened
Over the past several months, at least half a dozen U.S. federal district judges have issued standing orders that prohibit parties from using generative AI tools that are “public” or “consumer-grade” when handling discovery materials. Discovery is the phase where both sides exchange evidence—documents, emails, financial records, often containing highly confidential data.
The Bloomberg Law report that broke this story noted that judges are zeroing in on how public AI systems retain user inputs. Many free or consumer-tier AI tools use the data you submit to improve their models, store conversation histories, or share information with third-party service providers. That creates a clear risk: a lawyer uploading a client’s trade secrets or medical records into ChatGPT could inadvertently expose that data.
One order cited explicitly says that counsel “shall not input any confidential, proprietary, or sensitive information into any generative artificial intelligence tool.” Another requires attorneys to certify that any AI used in discovery does not store or train on the data submitted.
Why it matters
This is not an obscure legal technicality. The same mechanics that worry judges are at play every time you paste a document into a public AI chatbot. Here’s how it works:
- Data retention: Most consumer AI tools log your inputs. Even if the company says it “anonymizes” data, there is often a period where the raw text is stored.
- Model training: Some providers use your conversations to improve their models unless you explicitly opt out. OpenAI, for example, allows you to disable chat history, but the setting is not on by default.
- Third-party access: Chatbot providers sometimes subcontract to other companies (e.g., for content moderation) who could see your data.
The legal discovery context makes this risk acute because court rules require parties to protect privileged or confidential information. But the underlying vulnerability is universal. If you use public AI to draft a sensitive email about a medical condition, a business negotiation, or a personal dispute, you are effectively handing that text to a third party with no guarantee of privacy.
What you can do
You don’t have to stop using AI—but you should adjust how you use it when the information is sensitive.
Choose enterprise or private AI versions if you have access. Tools like ChatGPT Enterprise, Microsoft Copilot with commercial data protection, or Google Vertex AI offer contractual guarantees that your data will not be used for training. Many employers already have these licenses.
Turn off chat history and training in consumer tools. In ChatGPT, go to Settings → Data Controls → Disable “Chat History & Training.” This reduces retention, but check the provider’s privacy policy to confirm.
Never paste raw confidential information into a public chat interface. Treat the input box like a postcard—assume someone else might read it. Summarize or anonymize before using AI.
Read the privacy policy for each tool you use. Look specifically for language about “data usage for model training” and “sharing with third parties.” Policies can change, so revisit them periodically.
Consider local AI models that run entirely on your own device, such as Llama or Mistral via tools like Ollama. These never phone home.
If you are a professional handling client data, treat public AI the same way you would treat a cloud storage service without a business agreement. Get written consent or use a vetted enterprise tool.
Sources
Bloomberg Law News: “Judges’ Public AI Bans During Discovery Zero in on Privacy Risk” (June 5, 2026). Link to article (Note: RSS feed link, may not be directly accessible; search for the title on Bloomberg Law.)
OpenAI privacy policy and data controls. (Relevant sections available at openai.com/policies)
Standing orders referenced in the Bloomberg Law report are not publicly indexed but can be requested from the respective federal courts.
The takeaway is simple: the same privacy risk that has judges banning public AI in court is present every time you open a chatbot. Act accordingly.