Are My Employees’ AI Conversations Discoverable in a Lawsuit? 

Dear Will & AiME, 

Our employees increasingly use AI tools to brainstorm, summarize documents, work through customer issues, and think about business problems. If we end up in litigation, could those AI conversations have to be preserved or produced just like emails or text messages?  

–General Counsel, Cleveland 

Short Answer 💡

Yes. AI prompts, responses, uploaded files, and chat histories may be discoverable if they are relevant to a dispute and within the company’s possession, custody, or control. Businesses should account for AI conversations in their retention, litigation-hold, and privilege practices so relevant information can be preserved and sensitive communications handled appropriately. 

Dear General Counsel, 

AI tools are now part of everyday business communication, which gives companies a timely opportunity to bring prompts, responses, uploaded documents, and chat histories into their information-governance strategy. 

In the right circumstances, these conversations are business records, and in litigation they can be evidence. 

Not every AI conversation will be discoverable. Companies that plan ahead will know which ones matter and where to find them. 

Are AI Chats Electronically Stored Information (ESI)? 

Modern litigation routinely involves electronically stored information (ESI), including email, text messages, collaboration platforms, cloud documents, and internal messaging. AI conversations fit naturally within that framework. 

When an employee uses AI to analyze a dispute, draft a customer response, evaluate a product failure, summarize an internal investigation, or discuss a significant business event, the conversation may contain information relevant to a later lawsuit. 

The key question is whether the information exists, is within the company’s possession, custody, or control, and is relevant to the dispute. 

How AI Retention Settings Affect Litigation Holds 

Retention settings are where companies gain the most control. AI platforms handle retention differently. Some preserve chat histories, some let users delete them, and enterprise products often offer administrative controls, retention periods, and export tools that consumer accounts lack. 

Understanding those settings now puts companies in a stronger position later. Once a company reasonably anticipates litigation, it may need to preserve relevant information. If employees use AI in their work, litigation-hold procedures should account for AI conversations. 

Useful questions to ask: 

  • Where are the conversations stored? 

  • Can the company retrieve them? 

  • Can employees delete them? 

  • Are business users working through personal AI accounts? 

  • Does the company’s enterprise AI platform preserve prompts, responses, and uploaded files? 

Answering these questions early makes discovery faster, more efficient, and more predictable. 

Are AI Conversations Protected by Attorney-Client Privilege? 

Employees may use AI to ask about legal risk, summarize communications from lawyers, analyze disputes, or seek something resembling legal advice. Whether privilege applies depends on how the tool is used. 

Privilege generally covers confidential communications between lawyer and client for the purpose of legal advice. Adding a third-party AI service can complicate that analysis, particularly when counsel is not directing the tool’s use or the platform’s terms allow broader use or retention of the information. 

An AI tool used by outside counsel within a controlled legal workflow presents very different privilege issues from an employee pasting a lawyer’s confidential advice into a public chatbot and asking, “What does this mean?” 

Routing legal questions through enterprise AI tools directed by counsel is the best way to preserve existing protections. 

How Employee AI Prompting Habits Shape the Record 

People often write to AI more casually than they write formal business documents. An employee may type: 

  • “Why do you think this project failed?” 

  • “Help me explain why we missed the deadline.” 

  • “What arguments could the customer make against us?” 

  • “Rewrite this so it doesn’t sound like we knew about the problem.” 

Prompts like these feel informal and exploratory. In litigation, a prompt can show what an employee knew, what concerns existed at the time, or how the company was evaluating a problem. The employee’s question can matter as much as the AI’s answer. 

Practical Steps to Manage AI Chats in Discovery 

The most effective approach is to understand how employees use AI and build it into existing information-governance processes. 

Companies can update acceptable-use policies to address personal AI accounts, confirm that enterprise tools offer the right retention controls, and expressly include relevant AI chats and uploaded materials in litigation-hold procedures. 

Legal teams can also train employees to treat AI conversations with the same care as email. 

The familiar email rule applies here too: If you create a business record, someone may eventually ask to see it. 

Bottom Line: Treat AI Chats as Potential Business Records 

AI tools are now another place where employees think, communicate, and make decisions. Companies that treat key AI conversations as business records are better prepared for litigation and better governed day to day. 

Focus on where those conversations live, how long they are retained, whether they can be preserved, and whether employees are using AI for sensitive legal or business discussions that deserve extra care. 

-Will & AiME 

Three Takeaways: 

  1. AI prompts, responses, and uploaded materials may become discoverable electronically stored information when they are relevant to a dispute. 

  1. Understanding how AI platforms retain, delete, and export conversations before a preservation obligation arises gives businesses a strategic advantage. 

  1. Routing legal questions through enterprise AI tools directed by counsel helps preserve privilege and confidentiality. 


Next
Next

Can AI Help Me Search a New Logo Before We Launch It?