Discovery, but make it algorithmic. For lawyers, this is a warning flare. For Shell critics, it is something more delicious: a spectacle in which the oil giant, so often accused by campaigners of burying difficult truths in corporate fog, is now demanding transparency from the other side’s algorithmic workflow.
Shell, previously known as Forthdeal Limited, subsequently as Royal Dutch Shell plc, and now Shell plc after ditching the Royal Dutch name, has found itself orbiting yet another fascinating legal controversy — this time not over barrels, blowouts, reserves, pipelines, methane, LNG, Nigeria, Groningen, Prelude, or the usual industrial bingo card, but over something rather more modern: artificial intelligence prompts.
Yes, welcome to the age of climate litigation where even the questions typed into an AI system may become battlefield evidence.
The trigger is a recent Dechert article about a Connecticut federal magistrate judge’s order in Conservation Law Foundation, Inc. v. Shell Oil Company, et al., a case involving Shell-related defendants and allegations concerning climate-risk preparedness at a petroleum storage terminal. On 18 May 2026, Magistrate Judge Thomas O. Farrish ordered the plaintiff to produce generative AI prompts used by expert witness Dr Naomi Oreskes in preparing her expert report. Dechert notes that this appears to be the first federal court decision requiring an expert witness to disclose AI prompts as part of discoverable methodology — although the district court has stayed the order while the plaintiff’s objection is resolved. (Dechert)
That last caveat matters. Shell has not secured some final, thunderous legal victory in which the judiciary has declared all AI prompts fair game forever. The order is currently stayed. The dispute remains live. But the legal signal is still unmistakable: if an expert uses AI to filter, select, organise, or analyse evidence, the other side may argue that the prompts are not private doodles but part of the expert’s methodology.
In plain English: if your expert asks the machine to help sift the evidence, opposing counsel may ask to see exactly what the expert asked the machine.
And Shell, naturally, appears to have noticed the opportunity.
The underlying case is not some abstract academic seminar about robot lawyers. It is climate-adaptation litigation. According to analysis of the case, Conservation Law Foundation alleges that Shell and other operators of a bulk petroleum storage terminal in New Haven, Connecticut failed to account for climate-change-related risks such as sea-level rise, storm surge, and severe weather in the operation of the facility. The case has been described as one of the early U.S. climate adaptation cases to move deeply into discovery. (minerva26.com)
That is what makes the AI-prompt skirmish so revealing. Shell is not merely defending itself on the environmental merits. It is fighting over the architecture of the plaintiff’s expert evidence: how documents were selected, how the expert’s team narrowed the universe of Shell-produced material, and whether generative AI formed part of that process.
According to Mayer Brown’s summary, Dr Oreskes and a research assistant used a commercially available generative AI tool to filter and identify potentially relevant documents from defendants’ document productions. Shell and co-defendants later sought further information about the AI prompts and outputs, eventually moving to compel. Magistrate Judge Farrish found that expert methodology was legitimate ground for discovery and that the AI-assisted culling process was an aspect of that methodology. (Mayer Brown)
There is a certain grim poetry in this. Fossil-fuel litigation has entered the artificial-intelligence era, and Shell — a company that has long mastered the art of document production, corporate opacity, and reputational trench warfare — now wants to inspect the plaintiff expert’s prompts.
The corporate message is effectively: “We may have produced the documents, but we would now like to inspect the intellectual plumbing by which your expert found the embarrassing ones.”
This is not necessarily improper. In litigation, methodology matters. Expert witnesses do not get to wave a wand, mutter “AI helped,” and expect courts to treat the result as sacred scripture. If an expert uses a statistical model, sampling protocol, keyword search, or other method to select evidence, the opposing side will usually want to test it. AI prompts may, depending on their use, be the new search terms.
The problem is that generative AI is not merely a filing cabinet with a search bar. Prompts can reveal assumptions, strategy, emphasis, hypothesis, and sometimes counsel-influenced framing. A prompt can be a neutral retrieval instruction — or it can be a miniature theory of the case wearing a trench coat. That is why the fight matters.
Dechert’s article correctly identifies the broader risk: if this order survives, it could expand expert discovery by treating AI prompts as discoverable methodology rather than protected preliminary research. That could lead to demands for prompt histories, iterative refinements, system instructions, outputs, chat logs, and the entire messy sausage factory of AI-assisted expert work. (Mayer Brown)
For lawyers, this is a warning flare. For Shell critics, it is something more delicious: a spectacle in which the oil giant, so often accused by campaigners of burying difficult truths in corporate fog, is now demanding transparency from the other side’s algorithmic workflow.
There is a perfectly respectable legal argument behind Shell’s position. If an expert uses AI to narrow a document set, then the opposing party should be able to test whether the narrowing was reliable, biased, incomplete, or prompt-driven. Arnold & Porter’s commentary says the court treated the prompts as part of the expert’s methodology and noted that agreements protecting “notes” or “drafts” may not be enough unless AI materials are expressly covered. (Arnold & Porter)
But there is also an obvious strategic dimension. Climate litigation increasingly turns on internal documents, historical knowledge, corporate warnings, and what fossil-fuel companies knew or should have known. Dr Oreskes is not just any expert; she is a prominent historian of science associated with work on the manufacture of doubt. In a Shell climate case, that makes her evidence potentially uncomfortable. So Shell’s attempt to interrogate the AI route by which documents were selected is not merely a procedural sideshow. It is a fight over the lens through which Shell’s own record is examined.
One might almost admire the symmetry. Shell, whose public image has often depended on carefully polished messaging, now wants to know whether the other side’s expert used carefully framed prompts. Corporate communications meets courtroom computation. The washing machine is suing the spin cycle.
The unresolved procedural issues are important. Minerva26’s analysis notes that CLF has argued, among other things, that Shell did not serve formal Rule 33 interrogatories or Rule 34 requests for production covering the AI prompts, and that the district judge could potentially vacate the order on procedural grounds without deciding the broader substantive issue. (minerva26.com)
So the careful conclusion is this: the case has not yet created a final, settled rule that all expert AI prompts must be produced. But it has put every litigation team on notice. If AI touches expert evidence, the prompts may become discoverable. If lawyers want them protected, they had better say so clearly in discovery agreements. And if experts use AI casually, without preserving records or understanding the implications, they may be creating a future cross-examination exhibit.
For Shell plc’s investors, including large institutional names frequently listed among major holders such as BlackRock, FMR, Vanguard, and Norges Bank Investment Management, the story is another reminder that modern Shell risk is not confined to commodity prices and production volumes. Reported shareholder-data services currently list institutions as owning a substantial majority of Shell’s shares, with BlackRock, FMR, Vanguard, and Norges Bank Investment Management among the larger reported holders. (Simply Wall St)
The litigation risk is now reputational, environmental, technological, and procedural. It includes not only what Shell did, what Shell knew, and what Shell disclosed, but now also what an opposing expert typed into an AI tool while reviewing Shell’s documents.
This is the modern courtroom: oil terminals, climate science, expert reports, discovery sanctions, and prompt logs. Somewhere, a 20th-century oil baron is surely asking what happened to the good old days when a company merely had to worry about leaks from pipelines rather than leaks from chat histories.
The irony is rich enough to require its own carbon offset.
Shell and its defenders will say this is ordinary litigation discipline: test the expert, test the method, test the evidence. They have a point. Climate plaintiffs and their supporters will say this is a tactical attempt to burden, distract, and intimidate those trying to hold fossil-fuel infrastructure operators accountable for climate risks. They may also have a point.
Both things can be true. Litigation is not a morality play. It is a knife fight with filing deadlines.
But the larger lesson is difficult to miss. AI has now entered the evidentiary bloodstream. The prompt is becoming part of the paper trail. The question asked of the machine may be treated as part of the method. And in a case involving Shell, climate risk, and expert testimony, the fight over prompts has become a preview of where corporate litigation is heading.
For once, Shell is not drilling beneath the seabed.
It is drilling into the prompt box.
And the rest of us may soon discover whether the most dangerous document in climate litigation is no longer the internal memo, the board paper, or the risk report — but the little typed instruction that tells an AI system where to look.
Sir Henri Deterding and John Donovan mini-skit
Deterding: Ach, Mr Donovan, in my day we drilled for oil. Now they drill for prompts. This, I do not approve — though I admit, it is efficient.
Donovan: Shell appears rather keen on transparency when it is someone else’s AI workflow.
Deterding: Ja. A magnificent corporate principle: “Open the other fellow’s files, lock your own cabinet.”
Donovan: The court may yet decide the order went too far.
Deterding: Naturally. Lawyers first create fog, then invoice for lanterns.
Donovan: Still, it is a striking development. AI prompts may become part of expert methodology.
Deterding: Mein Gott. Even the machines must now prepare for cross-examination.
Donovan: Shell might call that accountability.
Deterding: Shell calls many things accountability. Sometimes it is merely a press release wearing a necktie.
DISCLAIMER: This article is opinion and commentary. It is based on cited public sources and should not be read as legal advice, financial advice, investment advice, or a definitive statement of the merits of any ongoing litigation. No allegation of unlawful conduct is made beyond what is contained in cited legal and journalistic sources. Site wide disclaimer also applies.
*This website and sisters royaldutchshellgroup.com, shellnazihistory.com, royaldutchshell.website, johndonovan.website, shellnews.net, and shellwikipedia.com, are owned by John Donovan - more information here. There is also a Wikipedia segment, the Shell DPA Files, "Shell and the Spies", the Shell Leaks files, as well as books written and published by John Donovan - Kindle eBooks. Timeline of the Donovan Shell Feud. Toxic History of Royal Dutch Shell Group. Shell and the Donovans: The Full Media Record — 550+ Articles, 110 Books, 40 Years. *All created and supported by internet wizz, Nick Gill.
