TL;DR:
- On November 4, 2026, the United States District Court for the Eastern District of California entered a Stipulated Protective Order As Modified in Estate of Ezekiel Rivera, et al. v. Kern County Sheriff's Department, et al. that bars the submission of confidential discovery material to open, public generative AI tools. The order allows use of enterprise AI tools that meet specified security and data practices. This marks a concrete, timely step toward AI-specific discovery protections in a live federal case and signals practical adjustments for trial teams handling sensitive data during discovery, preparation, and potential trial. (leagle.com)
Background and the action taken
Estate of Ezekiel Rivera v. Kern County Sheriff’s Department, et al. is an Eastern District of California matter filed in 2024, involving claims against Kern County and its sheriff’s department along with several deputies. On November 4, 2026, the court entered a stipulation for a protective order that, in its modified form, tightens the handling of confidential discovery materials and addresses AI usage directly. Several outlets summarizing the docket confirm the court entered Document 59—the stipulated protective order—as of that date, underscoring the court’s commitment to controlling AI-related disclosure in a high-stakes constitutional and civil action. The order is described as creating two protection tiers and, crucially, including a dedicated framework for using generative AI in the litigation. (leagle.com)
What the order actually does
Key provisions center on AI usage and the protection of confidential materials. The order distinguishes between Open Generative AI tools (public-facing systems like ChatGPT) and Permitted Generative AI tools (enterprise-grade systems with defined safeguards). Section 7.4, Use of Generative AI Tools, makes clear that CONFIDENTIAL material shall not be submitted to an Open Generative AI tool, and violations can trigger remedies under the protective order. Permitted tools are allowed only if they meet four criteria: (i) access is restricted to authorized users, (ii) the tool does not train public models on the protected data, (iii) there are retention and deletion controls, and (iv) security protections are commercially reasonable. The order thus codifies a concrete filter between consumer AI services and enterprise-grade platforms in the discovery process. This is not a mere policy recommendation; it is a court-ordered framework enforceable in this case. (lexsummary.com)
In addition, the order contemplates a two-tier confidentiality structure, with defined handling for CONFIDENTIAL and HIGHLY CONFIDENTIAL – ATTORNEYS’ EYES ONLY information, and it specifies that protected material may be used only for purposes contemplated by the litigation. The framework reflects a broader, growing trend in federal courts to regulate AI-assisted review and AI-driven handling of sensitive discovery while preserving access to technology that can improve efficiency under strict governance. (lexsummary.com)
Practical implications for trial teams
- Avoid public AI tools for confidential data. The explicit prohibition on submitting CONFIDENTIAL materials to Open Generative AI tools means litigation teams must audit what data they upload to AI platforms. Using public AI tools to summarize or process sensitive documents could constitute an unauthorized disclosure under the protective order, with consequences that could include sanctions. (lexsummary.com)
- Rely on enterprise AI within strict controls. Permitted tools offer the possibility of AI-assisted work in discovery and review, provided they meet the four criteria above. Firms should map which enterprise tools they will rely on, confirm vendor terms meet the protective-order benchmarks, and implement internal access controls to ensure only appropriate personnel can use the tools on protected data. (lexsummary.com)
- Structure and retention matter. The protective order envisions structured handling of confidential material, including retention and deletion protocols and court-supervised disposal after final disposition. This affects how teams store AI-processed outputs, how long copies exist, and how data is purged or archived. Practitioners should build a workflow that aligns with these safeguards to avoid risk of noncompliance. (lexsummary.com)
- Integrate AI governance into discovery planning. Given the order’s explicit treatment of AI tools, counsel should incorporate AI governance into Rule 26 disclosures, protective-order stipulations, and discovery plans. This includes briefing on what is permissible, who can access materials, and how AI-assisted tasks will be documented and reviewed for accuracy and relevance. (lexsummary.com)
- Prepare for potential sanctions or remedies. The order, like many protective orders with AI provisions, preserves the court’s authority to sanction violations. This heightens the need for diligent compliance in the discovery phase and during subsequent motions or hearings related to AI use. (lexsummary.com)
How to implement these protections in practice
- Conduct an AI-use inventory. Identify all stages where AI could touch confidential discovery data (e.g., review, summarization, translation, or extraction of data from documents) and categorize tools as Open AI vs Permitted AI. Ensure that any Open AI usage is eliminated for confidential materials.
- Choose compliant tools. For any AI usage, select enterprise-grade tools that meet the order’s four criteria, verify their data-handling policies, and ensure they do not train on the protected materials. Document vendor certifications and retention/deletion policies as part of the discovery plan.
- Train teams on proper use. Develop internal guidelines and training for attorneys, paralegists, and staff about what constitutes confidential material, which tools are permitted, and how to handle AI-generated outputs. Emphasize the need to verify outputs generated or processed by AI and to maintain a robust audit trail.
- Update protective orders in similar matters. Counsel should consider negotiating similar AI-specific protections in other cases, particularly when confidential discovery is involved or when high-sensitivity data is at stake. The Rivera protective order serves as a concrete template showing how AI restrictions can be embedded into court orders.
- Align with trial-readiness training. While the primary focus here is discovery governance, the trend toward AI controls dovetails with trial-readiness objectives. Objection-focused training and live simulations can help trial teams anticipate and address objections to AI-generated content or to alleged mishandling of AI tools during discovery and trial.
Objection Academy in the current AI risk landscape
As courts refine how AI is deployed in litigation, practical objection training becomes more important. Resources that drill objection-raising techniques, authentication challenges for AI-generated outputs, and strategies for preserving the integrity of evidence in AI-assisted workflows remain relevant for trial teams. In the wake of decisions like the Rivera protective order, structured practice in objecting to AI-driven evidence, ensuring proper foundation, and maintaining rigorous chain-of-custody for AI-processed materials can be decisive in trials and appeals.
Why this matters for trial attorneys now
The November 4, 2026 protective-order development in Rivera v. Kern County demonstrates that AI governance is moving from a theoretical policy debate into concrete, enforceable court-ordered rules in ongoing federal litigation. For trial teams, this translates into:
- Clear guardrails on what AI can and cannot do with confidential discovery data.
- A practical path to employing AI tools responsibly without triggering sanctions.
- A governance blueprint that can be adapted to other cases with sensitive information.
This development also reinforces the value of disciplined, practice-ready trial skills. Objection-driven training and evidence-readiness programs provide actionable techniques to navigate AI-influenced workflows in court, ensuring that the use of AI remains a strength rather than a vulnerability for trial teams.
Next steps for practitioners
- If handling a case with confidential discovery, review the Rivera order closely and assess whether similar protections are appropriate or already in place in the jurisdiction.
- Update internal discovery protocols to reflect AI restrictions, including the designation of permissible tools, data-handling procedures, and deletion/retention schedules.
- Develop or update an AI governance checklist for case teams, including training on how to preserve privilege, protect work product, and authenticate AI-generated outputs.
- Consider leveraging targeted objection training to prepare for challenges involving AI-derived materials at deposition, in motions, and at trial.
Sources:
- Estate of Ezekiel Rivera v. Kern County Sheriff’s Department, et al., No. 1:24-cv-00176-JLT-CDB, ED California, Stipulated Protective Order As Modified (Document 59) entered November 4, 2026. Leagle confirms the November 4, 2026 date and the order’s protective framework. (leagle.com)
- Estate of Ezekiel Rivera v. Kern County Sheriff’s Department, et al., ED California, Case No. 1:24-cv-00176, Justia docket entry reflecting the protective-order language and related discovery provisions (Document 59). (docs.justia.com)
- LexSummary overview of the protective order in Rivera v. Kern County, including the Open vs Permitted Generative AI Tool distinction and Section 7.4 Use of Generative AI Tools. (lexsummary.com)
Note: The Rivera protective-order development illustrates the practical, day-to-day impact of AI governance on discovery in federal litigation and provides a concrete model for trial teams seeking to reduce AI risk while preserving access to beneficial tools under controlled conditions.