Back to Articles

NVIDIA Discovery Stay in Rogers v. Nvidia Corporation Highlights Strategic Pause in AI Training Data Litigation

TL;DR:

In Rogers v. Nvidia Corporation, the Northern District of Illinois granted NVIDIA’s motion to stay discovery pending the court’s ruling on NVIDIA’s motion to dismiss. Decided September 24, 2026, the Memorandum Opinion and Order (MDO) held that a stay is appropriate to avoid burdensome, expensive discovery on multiple AI models while threshold issues are unresolved. The court allowed a narrowed scope of discovery focused on training datasets and data-field manifests, effectively delaying broad ESI and model-discovery while dispositive issues are briefed. The decision illustrates how courts are balancing the heavy burden of AI training data disputes against preservation duties, and it provides practical playbooks for defense and plaintiff teams as they navigate biometric privacy claims under BIPA and related theories. For trial teams, the ruling signals a strategic window to refine arguments, preserve resources, and recalibrate discovery plans. Objection Academy can help teams practice objections and trial-readiness for AI-data disputes and protective-order battles, ensuring readiness once discovery resumes.

What happened

On September 24, 2026 the United States District Court for the Northern District of Illinois issued a Memorandum Opinion and Order in Rogers et al v. Nvidia Corporation, holding that NVIDIA’s motion to stay discovery should be granted pending disposition of its motion to dismiss. The case, filed May 12, 2026, involves claims arising under Illinois Biometric Information Privacy Act (BIPA) and related statutes, based on allegations that Nvidia used journalists’ and voice actors’ voices to train AI models without consent. The court expressly recognized the substantial burdens associated with broad discovery across multiple AI models and datasets and concluded that a stay would conserve resources and avoid premature resolution of the case. The order cites standard stay-efficiency factors, including potential prejudice to plaintiffs, whether the stay would streamline issues, and whether discovery would be burdensome. The court determined all three factors weighed in favor of a stay, at least pending a ruling on the motion to dismiss. The stay covers discovery beyond narrowly defined categories, with plaintiffs limited to two specific requests: (1) documents identifying each accused AI model’s training datasets and dataset versions, and (2) documents describing the fields Nvidia maintains in its training data manifests. The decision and accompanying docket confirm that the judge sought to avoid a costly, sprawling discovery process that might be rendered moot by the threshold issues raised in the motion to dismiss. The MDO was signed by Judge Thomas M. Durkin and entered on September 24, 2026. (docs.justia.com)

Practical impact for trial teams

  • Discovery strategy pivots around dispositive issues. By staying broad discovery while NVIDIA’s motion to dismiss is pending, the court protects both sides from expending substantial resources on data collection that could be narrowed or eliminated by the court’s eventual ruling. For trial teams, this underscores the importance of sharpening threshold arguments early, particularly where AI training data, biometrics, and model training involve complex questions about consent, data provenance, and data-use restrictions. The Illinois court’s approach provides a blueprint: push for focused, state-law-aligned discovery that can be revisited if and when a decision on the motion to dismiss is reached.

  • Narrow discovery channels can still yield essential leverage. The court’s allowance of two narrow categories of documents—training datasets and the fields in training manifests—gives plaintiffs a clearly defined path to obtain critical information without triggering the full burdens of multi-model discovery. This is a useful template for future AI-data disputes, where parties often face unresolved issues about what constitutes relevant training data, how to trace data provenance, and what constitutes a permissible data field for model training.

  • Implications for biometric privacy cases. The case sits at the intersection of AI and biometric privacy law. The Seventh Circuit’s recent developments on BIPA amendments (and retroactivity) have clarified certain damages issues in biometric cases, and the NDIL stay here signals court willingness to coordinate lens on discovery with dispositive motions in this evolving field. Plaintiffs should monitor whether the stay extends or narrows as briefing on the motion to dismiss progresses, and defense teams should be prepared to pivot if the court narrows the scope of permissible discovery further or denies the dismissal.

  • Case management and scheduling considerations. For defendants, the ruling reinforces the strategic value of moving promptly on the threshold issues and suggests a disciplined approach to case management: propose targeted discovery plans, request protective orders as needed, and align discovery deadlines with anticipated court rulings. For plaintiffs, the decision encourages careful prioritization of what is most likely to survive a motion to dismiss and to focus responses accordingly, ensuring preservation remains robust while avoiding unnecessary data collection.

Practical next steps for litigators

  • If representing Nvidia or a similar AI-provider, anticipate a continued emphasis on threshold issues and consider narrow discovery plans that align with the court’s two-category approach. Prepare to brief the motion to dismiss thoroughly and to present a concise, model-specific plan for what discovery, if any, should proceed during the stay.

  • If representing plaintiffs, develop a precise argument for why limited training-data discovery is essential to liability. Prepare to expand discovery quickly if the motion to dismiss is denied or narrowed in a way that confirms liability theories. Consider offering a detailed data-provenance framework to assist the court in evaluating relevance and proportionality under FRCP 26.

  • Protective orders and data-handling provisions will be pivotal. In AI-related disputes, protective orders, confidentiality stipulations, and clear protocols for handling biometric data will be central to any discovery plan. Anticipate arguing for or against broad protective orders depending on whether the discovery relates to training data, model architecture, or training datasets.

  • Plan for the next procedural milestones. The case calendar will likely hinge on the disposition of the motion to dismiss. Prepare joint status reports, set realistic discovery cutoffs aligned with the dismissal timeline, and preemptively outline how to scale discovery if the court allows it in phases.

  • Consider training and advocacy tools. In this climate of AI-driven discovery disputes, trial-ready advocacy becomes critical. Objection Academy offers practical resources for trial teams to strengthen objection procedures, refine evidentiary arguments in AI data disputes, and rehearse protective-order negotiations. For teams facing similar cases, training in how to object to overly broad discovery demands and how to preserve privilege claims can improve courtroom readiness when discovery resumes.

Evergreen and practical notes

  • This development is part of a broader trajectory in AI and biometric-litigation discovery. Courts are actively evaluating how to balance the need for information with the burden of producing AI data, model training records, and biometric data. The September 2026 order in Rogers v Nvidia illustrates the judiciary’s pragmatic approach to conservation of resources while preserving the ability to litigate important issues.

  • For trial teams, the timing matters. A stay on broad discovery buys time to fortify dispositive motions, refine trial strategies, and align expert and lay-witness preparation with a narrower factual record. When discovery resumes, the team can leverage the narrowed scope to focus on the most consequential issues, including the provenance of training data and consent-based defenses.

  • Objection Academy remains a practical resource in this space. Training modules on evidentiary objections, deposition strategies, and trial-readiness can help trial teams prepare for the kinds of disputes that arise when AI training data and biometric data drive litigation. In the wake of a stay, a focused objection drill on scope limitations, data-protection protocols, and privilege issues can pay dividends at trial.

Sources

Note: This analysis summarizes publicly available court filings and reputable press coverage. The stay and its scope reflect the court’s order as of September 24, 2026, and subsequent developments may modify any aspect of discovery.

More Legal InsightsExplore MCLE Eligibility