TL;DR:
A July 2026 discovery ruling in Schulte v. LinkedIn Corp. confirms that a federal court will treat generative AI assisted document review as a conventional TAR process governed by reasonableness and proportionality, not with a special AI-only regime. LinkedIn’s use of Relativity aiR to make final responsiveness determinations was permitted, with the court applying standard Rule 26(b)(1) limits and declining sweeping demands for AI performance metrics. In a companion July 31 2026 order in Crowder v. LinkedIn, the court carved out a pragmatic compromise: LinkedIn must investigate and produce up to 100 hyperlinked documents produced by plaintiffs, rather than force broad, blanket AI disclosure. For trial teams, the message is clear: AI-enabled discovery is here, but it remains tethered to traditional discovery rules, proportionality, and carefully scoped requests. Practical steps below translate the ruling into courtroom-readiness.
What happened in Schulte v. LinkedIn
The Northern District of California became one of the first federal courts to adjudicate the use of a generative AI tool in discovery. On July 1, 2026, the court in Schulte v. LinkedIn Corp., No. 22-cv-00237-HSG (LB), approved LinkedIn’s use of Relativity aiR to filter and finalize document responsiveness, treating the technology as a TAR-like process subject to established discovery principles rather than a special regime for AI. The court emphasized proportionality and reasonableness under Federal Rule of Civil Procedure 26(b)(1), and notably declined to compel broader disclosures about the AI tool’s performance or internal operation absent a specific deficiency in production. This decision was widely reported as a landmark moment for litigants relying on GenAI in discovery and signaled that courts would scrutinize AI-assisted workflows through traditional discovery lenses rather than reinventing the wheel for AI alone. (bakermckenzie.com)
Less than a month later, on July 31, 2026, Magistrate Judge Laurel Beeler issued a companion discovery order in Crowder et al v. LinkedIn Corp., No. 4:22-cv-00237-HSG (LB). The order addresses a discovery letter brief and resolves questions about hyperlinked documents within produced emails. The court held that LinkedIn must investigate and produce, to the extent available, up to 100 hyperlinked documents identified by plaintiffs, rather than undertaking an expansive, burdensome search of every possible hyperlinked item. If a document has already been produced or withheld as privileged, LinkedIn must identify it by Bates or privilege log numbers; and LinkedIn must confirm in writing when a document is not available for production. The ruling reaffirmed that AI-assisted steps in discovery must still be tethered to proportionality and to the case’s actual needs. The court acknowledged the burden of hyperlinked discovery but found a middle-ground approach to be proportionate given the plaintiffs’ demonstrated relevance. (docs.justia.com)
Primary-source materials reflect the precision of these decisions. The July 31 order explicitly limits hyperlinked discovery and preserves the flexibility for the parties to negotiate further, while still ensuring targeted, relevant inquiry. A contemporaneous order from June 30, 2026 (signed July 1, 2026) in Schulte confirmed the initial acceptance of Relativity aiR for final responsiveness determinations. Together, these rulings establish a practical framework for AI-enabled discovery that many trial teams will encounter in the near term. (docs.justia.com)
Practical implications for trial teams
AI-enabled discovery is permissible, but courts will apply conventional standards. The Schulte decision underscores that GenAI tools used in document review do not escape the basic duties of proportionality, relevance, and reasonableness under Rule 26(b)(1). Attorneys should frame AI workflows within these traditional parameters and resist any assumption that AI requires a special, higher level of scrutiny or disclosure absent a deficiency. (bakermckenzie.com)
Limit the scope of AI-related requests to what is actually needed. The Crowder order demonstrates that courts are skeptical of sweeping AI-related investigations or punitive disclosures about internal AI metrics unless a party can show a concrete deficiency. Instead, courts may accept a targeted, capped approach—here, up to 100 hyperlinked documents identified by plaintiffs. This matters for trial teams when negotiating ESI orders and discovery plans, as it provides a workable template for AI-assisted searches without collapsing into an overbroad, unmanageable data dump. (docs.justia.com)
Balance AI tool disclosures with privacy and burden concerns. The Crowder decision emphasizes that requests to reveal every social media connection or other personal data can be weighed against privacy burdens and proportionality. Absent a strong showing of relevance, courts may limit or deny such sweeping requests, even when AI tools are employed to process the data. Legal teams should craft discovery requests that privilege precision and relevance, and rely on proportionality to prevent disclosure creep. (docs.justia.com)
Expect structured cooperation around AI workflows. The orders show a preference for parties to meet and confer and to implement pragmatic, scaled approaches rather than unilateral, intrusive demands. In Schulte and Crowder, the courts encourage a cooperative process to define how AI will be used, what populations will be reviewed, and how results will be validated and logged. Trial teams should plan for early, precise ESI protocols that explicitly address AI usage, sampling, validation, and QA processes. (docs.justia.com)
Use AI-savvy discovery language in motions and orders. As courts increasingly accept AI-assisted review, it becomes important to include explicit language in ESI orders that contemplates TAR tools, the level of human review, sampling procedures, and the proportional limits on AI-driven productions. This can prevent later disputes and set up a clear playbook for ongoing litigation. The Schulte/Crowder line provides a concrete blueprint for such language. (bakermckenzie.com)
Prepare for cross-examination and trial-readiness on AI outputs. With AI already playing a role behind the scenes in discovery, trial teams should be ready to address the reliability and relevance of AI-derived outputs during trial. Objection Academy’s resources on evidence training and objection drills can help teams anticipate objections to AI-generated outputs and ensure a rigorous, courtroom-ready approach to GenAI-driven documentary evidence and summaries. The emerging practice is to couple disciplined AI workflows with robust issue-spotting and objection preparation at trial. (Objection Academy remains a practical resource for objections to AI-assisted evidence, and for building a repeatable, trial-ready approach to complex eDiscovery narratives.)
How Objection Academy fits into this new discovery landscape
As AI becomes a standard part of discovery workflows, trial teams benefit from structured practice on objections to AI-derived evidence, chain-of-custody concerns for AI outputs, and the thorough vetting of AI-assisted conclusions before they reach the jury. Objection Academy offers objection drills and evidence training that help teams anticipate and counter AI-related challenges in both pretrial and trial settings. By integrating these practices with a disciplined AI-enabled discovery plan, litigators can maintain credibility with the court while leveraging the efficiency of AI tools.
Concrete next steps for the litigation team
- Review the Schulte and Crowder orders to align your internal Discovery Plan with the proportionality standards that federal courts are already applying to AI-assisted review.
- When negotiating ESI orders, propose a concrete cap and a precise process for AI review, including the use of human quality control sampling and a defined timeline for any AI-driven productions.
- Draft a stand-alone section in your discovery plan that addresses Relativity aiR or any GenAI TAR tool, including data security, access controls, and the scope of review population.
- Prepare a brief, focused cross-examination outline for cases involving AI-derived evidence, drawing on Objection Academy’s training materials to craft appropriate objections and qualification questions.
Sources
Kevin Schulte, et al. v. LinkedIn Corporation, No. 4:22-cv-00237-HSG (LB), Discovery Order, ECF Nos. 186, 188, 190; Signed June 30, 2026; Filed July 1, 2026. Comprehensive orders and discussion of Relativity aiR for final responsiveness determinations. (docs.justia.com)
Schulte v LinkedIn Corp., 2026 WL 1905851 (N.D. Cal. July 1, 2026). First federal decision approving generative AI for document review under traditional TAR standards. (bakermckenzie.com)
Crowder et al v. LinkedIn Corporation, No. 4:22-cv-00237-HSG (LB), Discovery Order addressing 220 Discovery Letter Brief, Dated July 31, 2026. Up to 100 hyperlinked documents production cap; specific meet-and-confer framework. (docs.justia.com)
Court coverage and analysis from law firms highlighting the practical implications for discovery practice, including the emphasis on proportionality and AI workflow integration. (dlapiper.com)
Centered discussion on AI in discovery from industry observers and relevant commentary on AI-evidence governance. (uscourts.gov)
Advisory material on the emergence of AI rules and AI-evidence considerations in federal practice, including public-facing summaries of Rule 707 discussions. (uscourts.gov)
End of piece.