Artificial intelligence is no longer a future-state consideration for legal teams; it is an active, embedded part of how organizations operate today. Tools like ChatGPT and Microsoft Copilot are now routinely used across business functions, from drafting communications to analyzing data and supporting decision-making.
As adoption accelerates, a new question is emerging in litigation and investigations: Is AI-generated content, and the prompts behind it, discoverable?
The short answer is yes. The more nuanced answer is where things become interesting.
Courts and litigants are beginning to recognize AI chatbot interactions as a form of electronically stored information (ESI). In the case of Tremblay v. OpenAI, Inc., plaintiffs were granted the ability to request ChatGPT prompt data, marking an early but important signal that this category of data is not off-limits.
While still relatively uncommon in day-to-day discovery requests, this development reflects a broader trend: If AI tools are used in the ordinary course of business, their outputs and inputs may be relevant to a matter.
Unlike traditional data sources such as email or file shares, AI-generated content introduces new layers of complexity:
A standalone output may not tell the full story. The prompt that generated it can be equally, if not more, important in understanding intent, decision-making, or knowledge. Moreover, AI interactions are often iterative; users frequently refine or adjust their prompts based on previous responses, creating a dynamic dialogue rather than a one-off exchange. Each prompt and response builds upon the last, contributing to a layered context that may reveal the evolution of thought process, strategy, and information gathering. As a result, assessing just the final output without the sequence of interactions risks missing critical context and nuance embedded in the back-and-forth that characterizes many real-world uses of AI chatbots.
AI interactions may reside across multiple environments, standalone tools, enterprise platforms, or integrated systems within productivity suites. This fragmentation creates unique challenges for legal teams seeking to identify, preserve, and collect relevant data. Unlike traditional sources such as email servers or document repositories, AI-generated content can be scattered across a range of applications, some cloud-based, others locally deployed, and many embedded within broader business ecosystems. As organizations increasingly adopt specialized AI tools alongside mainstream platforms like Microsoft 365 or Google Workspace, the number of potential data sources grows, making a comprehensive discovery process more complex. Additionally, privacy controls, data retention policies, and user-level permissions can vary widely, impacting how easily chatbot interactions can be located and exported. Legal teams must therefore develop new strategies for mapping data flows, collaborating with IT, and ensuring that no relevant AI-generated evidence slips through the cracks during discovery.
AI platforms are continuously evolving, which means data formats, retention policies, and accessibility can vary significantly over time. This dynamic landscape draws a strong parallel to mobile device collections, where frequent software updates, shifting hardware standards, and changing app protocols create a moving target for forensic teams. Just as the forensic collection of mobile devices requires careful attention to evolving file types, metadata, and operating system nuances, so too does the collection of AI chatbot data. Vendors may introduce new export formats or adjust default retention periods without notice, complicating preservation and retrieval efforts. Permissions and access to historical interactions can shift, and integration with other business applications further increases complexity. In both cases, achieving a forensically sound collection demands proactive strategies: regularly monitoring platform or device updates, adapting eDiscovery protocols, and collaborating closely with IT to ensure no critical information is overlooked or rendered inaccessible. The challenges faced in mobile device forensic collection underscore the need for legal and technical teams to anticipate and address ongoing changes in AI platforms, so that collections remain robust and defensible.
From a technical standpoint, collecting AI chatbot data is not as disruptive as it may seem, provided you have the right tools and expertise.
For example, enterprise ecosystems are already adapting. AI interactions within platforms like Microsoft 365 can be collected through integrated compliance tools, where chatbot activity is captured alongside other user-generated content. Similarly, solutions like RelativityOne now support direct collection from sources, including ChatGPT, through dedicated connectors.
In many cases, chatbot interactions can be exported in structured or text-based formats and processed as standard documents within an eDiscovery workflow.
The takeaway: While the data source is new, the foundational workflows of collection, processing, review, and production remain familiar for now.
Even if AI chatbot data is not yet a routine request, legal teams should not treat it as a future problem.
Organizations are already using AI to:
That means relevant evidence may already exist within these tools.
Failing to account for AI-generated content introduces risk, including:
At Array, we view AI chatbot data as part of a broader evolution in the types of ESI organizations must manage, not as an outlier.
Our approach is grounded in three core principles:
We leverage a diverse ecosystem of tools to collect data from both traditional and emerging sources, including AI platforms, cloud applications, and collaboration tools.
From identification and preservation through production, our processes are designed to ensure consistency, transparency, and defensibility, regardless of data type.
Our team works closely with clients to determine when and how AI data should be incorporated into discovery, balancing risk, cost, and relevance.
Because most AI chatbot data can be normalized into standard document formats, we are able to integrate it seamlessly into existing review and analysis workflows, reducing disruption while maintaining rigor.
Today, requests for AI chatbot data may be limited. Tomorrow, they are likely to be expected. The rapid evolution of AI technologies is fundamentally reshaping organizational communications and operational processes. As businesses integrate chatbots and other AI-driven platforms, the resulting data is becoming an increasingly valuable and relevant asset within the legal landscape. Courts and regulatory bodies are beginning to recognize the importance of this information, which means that what is seen as an edge case now will soon become routine in discovery workflows. Proactive organizations that assess, prepare, and adapt will find themselves ahead of the curve, ready to respond confidently as expectations shift.
As AI adoption continues to expand, so too will its role in litigation. What we are seeing now is the early stage of a shift, one where AI-generated content becomes just another category of discoverable information. Legal teams that start preparing now, by identifying where AI data resides, understanding how it is used, and establishing defensible protocols for collection, will be better positioned to reduce risks, costs, and delays. Early investment in expertise and infrastructure ensures that organizations can seamlessly integrate AI content into eDiscovery processes, maintaining both rigor and responsiveness. Ultimately, being prepared transforms the challenge of AI data into a strategic advantage rather than a source of disruption.
AI is not just transforming daily operations; it is fundamentally redefining how organizations communicate, collaborate, and make decisions. As the world rapidly embraces AI for everything from automating routine tasks to driving strategic insights, its influence now extends far beyond the boundaries of traditional business processes. The implications for eDiscovery are profound: this discipline must not merely keep pace, but evolve in lockstep with the accelerating adoption of AI.
The question is no longer if AI chatbot data will become a central facet of discovery, but how prepared legal teams are to adapt as this technology takes its rightful place at the forefront. With courts and regulators increasingly recognizing the significance of AI-generated data, organizations that proactively assess, integrate, and manage this evolving evidence will be poised to turn the challenge of AI into a strategic advantage.