TL;DR
Three federal court rulings from earlier this year have redefined how attorneys must handle AI during discovery. Publicly accessible “open” AI platforms, where data is used to train the software, now pose a documented legal risk. As a result, courts are issuing protective orders restricting their use in all discovery materials, not just confidential ones. If your current protective order does not specify your choice of AI tool, it may not meet new judicial expectations.
The question used to be whether attorneys should use AI in discovery. These court cases show that we are way beyond that elementary question. The question now is which AI tools are legally permissible, and your protective order needs to answer it before a dispute arises.
Three federal decisions from early 2026 sketch out a framework that every litigation team should understand before the next case kicks off. The rulings come out of Michigan, Kansas, and Colorado, and they point in the same direction.
The Kansas Decision That Changes the Standard
In Jeffries v. Harcros Chemicals, Inc., the United States District Court for the District of Kansas granted a motion to amend a protective order to restrict the use of open or publicly accessible AI tools. The court applied that restriction not only to confidential information but to all discovery materials, including those that are not confidential.
The court’s most significant finding was that open, publicly accessible AI tools create an irreversible loss of control. Submitted data is incorporated into the platform’s ongoing model training, making it impossible to retrieve or delete after the fact. The court also found that submitting large volumes of discovery to a public AI repository carries legal risk even when individual documents are not technically confidential, including potential violations of U.S. and European data privacy laws.
The court noted that defendants’ employees, contractors, and correspondents had not consented to their data being submitted to a publicly accessible AI platform, and that a proposal to redact only personally identifiable information did not appear to comply with GDPR consent requirements. The court ordered that only closed AI tools could be used. That is the incredibly important standard this ruling sets.
What Colorado Added to the Framework
In Morgan v. V2X, Inc., from the District of Colorado, the court provided practitioners with the most operationally specific guidance yet on what an AI-aware protective order should contain. The court confirmed that AI-assisted litigation materials prepared with publicly available AI tools can be protected as mental impressions and litigation preparation materials under Rule 26(b)(3), and that using a consumer AI platform does not automatically waive that protection. Critically, the court required that any AI tool used to process confidential information be subject to key contractual safeguards, including prohibitions on using inputs for model training, restrictions on onward disclosure, and the ability to delete the data on request.
This provides you with a substantive, well-defined, and well-guarded checklist of AI tools that may be used for specific purposes. If you cannot verify that your AI vendor contractually prohibits training on your inputs, restricts onward sharing, and allows deletion on request, you do not have a defensible AI workflow in discovery.
Michigan Confirmed the Work Product Question
In Warner v. Gilbarco, Inc., from the Eastern District of Michigan, the Court extended work-product protection to AI-assisted litigation-preparation materials. Courts are willing to protect AI-assisted legal work when it reflects genuine litigation strategy, but the outcome turns on who used the tool, for what purpose, and under what safeguards. The consistent pattern across all three cases is that tool choice and data governance are what determine whether protection holds.
The court confirmed that a waiver of privilege over materials occurs when there is “voluntary disclosure…to a third party…” The court further stated that GenAI programs are “tools, not persons” and fall outside the scope of a waiver of privilege. The court also goes on to affirm that analysis, thoughts, and mental impressions are covered by the work product doctrine. In Jeffries, the protective order applied to both confidential and non-confidential discovery information, making it one of the broadest restrictions courts have put in writing so far. That breadth reflects a growing judicial consensus that AI platform selection is a discovery governance issue, worthy of express court control.
The Questions Your Discovery Team Should Be Asking Right Now
We want to give you the right questions to ask before any AI tool is used on your next matter:
- Does the platform retain your inputs and train on them?
- Does the vendor contractually guarantee no onward disclosure?
- Can you request the deletion of matter-specific data when the case closes?
- Does your current protective order address AI tool choice by category or by name?
If you cannot answer those questions confidently, you are working on an assumption. Assumptions do not survive discovery disputes and do not satisfy the Rules of Professional Responsibility.
AI Tool Selection Is a Discovery Management Decision
Choosing an AI tool for document review is a decision that may carry significant legal consequences. These three rulings make that clear. Your discovery partner should already be raising this with you, asking the questions that protect the case before a dispute surfaces.
Complete Legal can help you and your team choose private, secure models that maintain confidentiality and support the kind of defensible workflow courts are now expecting. When the opposing party challenges your process, the answer cannot be that you used whatever tool was available.
The courts in Michigan, Kansas, and Colorado have given the eDiscovery community a framework. The next step is to ensure your protective order, vendor contracts, and AI workflow align with that framework.
If you want to talk through how your current discovery process maps to these emerging standards, we are ready to have that conversation with you.
Contact Us:
info@completelegal.us | 
888-885-9909
Key Takeaways
- Courts are now granting protective orders restricting the use of open, publicly accessible AI tools in all discovery materials, including nonconfidential materials. Jeffries v. Harcros is the clearest ruling to date.
- The Morgan v. V2X decision provides practitioners with a concrete checklist: your AI vendor must prohibit training on inputs, restrict onward disclosure, and allow deletion upon request.
- AI tool selection is a legal strategy decision. Your protective order, your vendor contracts, and your discovery workflow need to reflect the standards courts will have already set in 2026.
