Are Ai Prompts Discoverable

Yes -- consumer AI conversations are discoverable, and the Heppner ruling made this explicit. The court found that conversations with consumer AI tools like ChatGPT are not protected by attorney-client privilege because the AI is not your client, your attorney, or a privileged intermediary. Enterprise AI platforms may have stronger protections under the Kovel doctrine, but the law is still developing and no court has definitively ruled that enterprise AI conversations are privileged.

This means every prompt you type into ChatGPT, every strategy discussion you have with Claude, and every document you feed into a consumer AI tool could end up in opposing counsel's hands during discovery. The implications for litigation strategy are enormous -- and most attorneys haven't thought about this at all.

The Heppner Ruling: What It Actually Decided

In Heppner, the court addressed whether conversations between an attorney and a consumer AI tool were protected from discovery. The ruling was unambiguous: they're not.

The court's reasoning was straightforward. Attorney-client privilege protects confidential communications between an attorney and their client for the purpose of obtaining legal advice. An AI chatbot is not a client, not an attorney, and not an agent of either. Sharing information with an AI tool is sharing information with a third party -- and voluntary disclosure to a third party generally waives privilege.

The ruling specifically addressed consumer AI platforms where the provider's terms of service may allow data to be stored, processed, or used for training. The court found that using such platforms was analogous to discussing case strategy in a public place -- you've voluntarily disclosed the information to a party with no confidentiality obligation to you.

Heppner didn't address every AI scenario. It focused on consumer tools with standard terms of service. But the reasoning has clear implications for any AI platform where the attorney can't demonstrate a confidentiality obligation from the AI provider.

Enterprise AI: The Kovel Doctrine Argument

Enterprise AI platforms present a different legal question. Under the Kovel doctrine (United States v. Kovel, 1961), communications with third-party experts retained to assist with legal representation can be privileged if the expert is necessary for the attorney to provide legal advice.

The argument for enterprise AI privilege runs like this: the law firm retains an enterprise AI platform (with a contractual confidentiality obligation) as a tool necessary for providing legal services. Communications with that tool are functionally equivalent to communications with a Kovel expert -- they're made in confidence, for the purpose of legal representation, and the third party has a contractual duty to maintain confidentiality.

No court has definitively accepted this argument yet. But the structural elements are stronger than the consumer AI case:

Contractual confidentiality: Enterprise agreements typically include data protection provisions, non-disclosure obligations, and commitments not to train on client data.

Necessity: If the AI tool is integral to the legal research and analysis process, it strengthens the argument that it's a necessary expert under Kovel.

Control: Enterprise platforms offer administrative controls, audit logs, and data governance that consumer tools don't -- demonstrating the firm's intent to maintain confidentiality.

The smart move is to structure your enterprise AI use as if the Kovel argument will succeed while planning for the possibility that it won't.

What's Discoverable and What Might Be Protected

Based on Heppner and the developing case law, here's the current landscape:

Clearly discoverable: - Conversations with ChatGPT Free, Plus, or any consumer AI tier - Prompts entered into any AI platform without contractual confidentiality protections - AI-generated outputs that were shared with third parties or filed with courts - AI conversation histories stored on the provider's servers under standard terms of service - Prompts and outputs from personal (non-firm) AI accounts used for client work

Potentially protected (untested): - Conversations with enterprise AI platforms under contractual confidentiality agreements - AI interactions within privileged attorney-client communications (e.g., AI embedded in secure client portal) - Prompts that reflect attorney mental impressions and work product (work product doctrine, not privilege) - AI-assisted analysis within a firm's internal work product systems

The work product angle: Even when privilege doesn't apply, the work product doctrine may protect some AI interactions. Attorney mental impressions, legal theories, and case strategy reflected in prompts could qualify as opinion work product -- which receives near-absolute protection. But factual work product (asking AI to summarize documents) receives only qualified protection and can be overcome with a showing of substantial need.

How to Structure AI Use for Maximum Protection

Given the discoverability landscape, attorneys should structure their AI use defensively:

1. Use enterprise platforms exclusively for client work. Consumer AI accounts should never be used for client matters. The cost difference between free ChatGPT and an enterprise legal AI platform is trivial compared to the discoverability risk.

2. Separate personal and professional AI use. If you use AI personally, keep it on a completely separate account from any professional use. Commingling personal and professional AI conversations creates discovery nightmares.

3. Treat every prompt as potentially discoverable. Before typing a prompt, ask: would I be comfortable with opposing counsel reading this? If not, don't type it. Don't include case strategy, attorney mental impressions, or privileged communications in AI prompts unless you're on an enterprise platform with Kovel-worthy protections.

4. Maintain prompt logs selectively. Some firms log all AI prompts for quality control. This creates a discoverable record. Consider whether comprehensive logging serves your interests or creates unnecessary risk. At minimum, develop a retention policy for AI interaction records.

5. Build the Kovel framework now. If your firm uses enterprise AI, document why the AI tool is necessary for providing legal services, ensure contractual confidentiality provisions are robust, and treat the AI platform as you would any Kovel expert. This positions you for the best possible argument if discoverability is challenged.

Implications for Litigation Strategy

AI discoverability creates new tactical considerations that litigators need to internalize:

Offensive discovery: You can now request opposing counsel's AI conversations in discovery. If you suspect opposing counsel used consumer AI to develop their arguments, a targeted discovery request for AI interaction records is legitimate. The Heppner framework supports it.

Defensive posture: Assume your AI interactions will be requested. This changes how you use AI in active litigation. Draft prompts that wouldn't embarrass you on cross-examination. Don't use AI to generate arguments you'd be uncomfortable defending as your own.

Impeachment potential: If opposing counsel submits a filing and their AI conversations show they knew about weaknesses in their argument (because the AI flagged them), those conversations could be used for impeachment. AI interactions are a new category of evidence that litigation practice hasn't fully reckoned with.

The metadata problem: AI platforms store metadata -- timestamps, session lengths, prompt histories. This metadata can reveal the timeline of legal research and strategy development. In cases where timing matters (when did counsel become aware of X?), AI metadata could be highly relevant.

The Bottom Line: Consumer AI conversations are discoverable under the Heppner ruling -- treat every prompt as if opposing counsel will read it, use enterprise platforms for client work, and structure your AI use for Kovel protection.