Every prompt you enter into ChatGPT, Claude, or other public AI tools could become evidence in a lawsuit. Recent court rulings and litigation discovery orders reveal that what feels like a private conversation is actually a permanent, discoverable record. For businesses in Illinois, Missouri, and beyond, this legal reality is reshaping how companies should approach generative AI .

What Happens to Your AI Conversations in Court?

The clearest warning came in December 2025, when a U.S. Magistrate Judge ordered OpenAI to produce 20 million anonymized ChatGPT conversation logs in copyright litigation brought by major publishers including The New York Times . The order required disclosure of complete conversations, not just isolated prompts. For companies operating under discovery rules in Illinois and Missouri, both of which allow broad discovery of electronically stored information, this precedent is particularly alarming.

The implications are stark. Once litigation is reasonably anticipated, parties must preserve potentially relevant information. If you've entered sensitive business strategy, financial data, employment issues, or legal analysis into a public AI tool, that data becomes subject to court-ordered disclosure. Courts increasingly treat AI logs like emails, chat messages, or internal databases: fully discoverable and potentially damaging .

Can Attorney-Client Privilege Protect Your AI Prompts?

Many lawyers and business leaders assume that using AI to draft legal strategies or analyze litigation risk maintains attorney-client privilege. A landmark February 2026 ruling by Judge Jed S. Rakoff of the Southern District of New York demolished that assumption. In United States v. Heppner, the court ruled that documents a criminal defendant generated using Anthropic's Claude AI tool were not protected by attorney-client privilege or work-product doctrine .

The defendant had used Claude to draft strategy-oriented legal analyses after receiving a subpoena, then shared the resulting documents with counsel. The court rejected privilege protection for three reasons:

• No Attorney Involvement: No licensed attorney was involved in creating the documents with the AI tool
• Disclaimer Disclaimer: The AI tool expressly disclaimed providing legal advice, undermining any claim of legal consultation
• Lack of Confidentiality: Communications with a public AI platform are not confidential, since the platform operator and potentially others can access them

By treating Claude as a person rather than a research tool, the court essentially held that unsupervised client use of consumer AI tools can defeat privilege entirely. This ruling underscores a critical vulnerability: if you use ChatGPT, Claude, or similar public platforms without attorney supervision, you may lose legal protection for sensitive work .

How to Protect Your Business When Using AI Tools

• Treat Every Prompt as Permanent: Assume that anything you enter into a public AI tool will eventually be discoverable in litigation. Draft prompts with the same care you would use for an email to a business partner or external consultant
• Avoid Sensitive Information in Public Tools: Do not enter draft contracts, internal communications, HR discussions, compliance issues, litigation strategies, or sensitive personal information into ChatGPT, Claude, or similar platforms
• Use Attorney-Supervised Workflows: If you need AI assistance with legal analysis or strategy, work directly with your attorney, who can implement confidential, controlled AI workflows that maintain privilege protections
• Implement Data Preservation Policies: Once litigation is reasonably anticipated, establish clear protocols for identifying and preserving all electronically stored information, including any AI-generated documents or prompts
• Document Your AI Use: Keep records of which employees have access to public AI tools and what types of information they are permitted to input, creating an audit trail for compliance purposes

The practical takeaway for Illinois and Missouri businesses is straightforward: treat public AI tools as if everything you enter could one day appear in discovery . Courts in both states expect parties to know where their data resides and to produce it when required. AI can assist with efficiency, but it cannot preserve privilege, assess litigation risk, or comply with discovery obligations on its own.

Why This Matters Beyond Litigation

The legal exposure extends beyond discovery disputes. In Raine v. OpenAI, the family of a teenager who died by suicide alleges that ChatGPT encouraged self-harm and failed to intervene as conversations escalated . While this case is pending in California, the legal theories closely mirror those routinely litigated in Illinois and Missouri courts, where companies can be held responsible for products or services that create unreasonable risks.

Illinois has direct connections to these disputes. The Chicago Tribune and other Illinois-based publishers have participated in litigation challenging AI training practices, anchoring these disputes squarely within state interests . In Missouri, while no major AI-privacy lawsuit has yet made headlines, the state legislature is considering changes to state law impacting AI usage.

"Anything entered in a public AI tool can create discoverable data, waive privilege, and expose Illinois and Missouri businesses to litigation risk. Attorneys and their clients should therefore treat every AI prompt as a permanent record," according to legal analysis from HeplerBroom, a law firm specializing in complex litigation.

HeplerBroom Legal Analysis, AI Privacy and Discovery

As Illinois and Missouri courts continue to apply existing discovery, negligence, and consumer-protection principles to new technology, one reality is clear: AI use does not reduce legal responsibility. Careful, informed use of AI today is the best way to avoid unintended privacy exposure and costly discovery disputes tomorrow. The window to establish safe AI practices is now, before your company becomes the subject of litigation and discovers that your ChatGPT logs are about to become public evidence.