Discovery’s New Frontier: Requesting the Plaintiff’s GenAI Data

Yesterday, I talked about the obligations that employers have to preserve data. But let’s switch gears; in an employment discrimination claim brought against a company, an employee’s AI usage is fair game for discovery.

And by not asking for it in litigation, you just might be missing out.

ChatGPT alone has over 700 million weekly users. Gemini just launched a new powerful version this week. And Claude is another favorite. Employees are using these tools to discuss workplace issues, research their legal rights, and draft complaints. These conversations are discoverable and may contain valuable impeachment material or evidence of the plaintiff’s true motivations.

So, when drafting discovery requests in employment cases, what can be asked for? Here are a few ideas:

• GenAI conversation history related to the claims. If the plaintiff used ChatGPT to draft their EEOC complaint or analyze whether they have a legal claim, those conversations are discoverable. They may reveal the plaintiff’s actual motivations, timeline of events, or inconsistencies with their stated claims.
• Prompts and responses about workplace issues. Employees may use AI to vent about workplace problems, seek advice on how to handle situations, or explore their legal options. These conversations provide context missing from formal complaints.
• AI-generated documents. If the plaintiff’s demand letter or complaint shows signs of AI assistance, request the underlying AI conversations. The prompts employees used often reveal more than the polished final product.
• Timeline of AI usage. When did the plaintiff start using AI to research employment claims? An employee researching “retaliation claims” even before making a complaint may have ulterior motives.

Employers should discuss with their counsel discovery requests broadly enough to capture AI-assisted communications. Traditional language requesting “emails and text messages regarding your claims” may not be specific enough to request AI data.

While courts are slow to catch up to this as part of mandatory initial disclosures, that doesn’t mean employers can’t include AI-related discovery in their initial requests. Many plaintiffs don’t realize their AI conversations are discoverable. Early requests capture data before the plaintiff understands what’s at stake.

Once plaintiffs learn their ChatGPT history is discoverable, they may delete conversations or claim they don’t exist. Early preservation demands and discovery requests reduce this risk.

It doesn’t take too much imagination to think about what employers might find in the responses to the Gen AI requests, things like:

• Admissions that workplace criticism was justified or performance issues were real.
• Evidence the plaintiff was planning to leave anyway or was already job searching before the alleged adverse action.
• Inconsistent timelines showing the plaintiff’s current version of events differs from their contemporaneous understanding.
• Proof the plaintiff manufactured or exaggerated claims after consulting with AI about what makes a strong legal case.
• Discussions with AI about how to maximize settlement value or what claims to add for leverage.

GenAI discovery is cutting edge, but with the technology moving quickly, the amount of data that should be available to parties just continues to increase.