Angus Menzies, Daniel Rubin & Mr B Gray : AI Scrutiny

Three men have inadvertently assembled a rather magnificent argument about the future of British justice. One expects an Employment Tribunal to return with improved reasons supporting its original decision. Another warns that organisations ignoring uncomfortable problems risk being mauled across social media. The third questions whether an unrepresented litigant using artificial intelligence understands the words submitted in his name.

Individually, their contributions have already received TCAP’s attention. Together, they reveal something considerably more interesting than another disagreement over courtroom conduct. Legal authority is becoming easier to examine. The practical information advantage enjoyed by legal professionals is diminishing. And the people accustomed to interpreting everybody else’s arguments are discovering that outsiders can scrutinise their words too.

Before anyone complains about the seating arrangements, our third contributor is a guest. Mr B Gray appeared for Page Outsourcing in my Employment Tribunal proceedings. He does not belong to Horsfield Menzies, whose Daniel Rubin represented Cepac. Angus Menzies did not appear at the hearing either. We’re renting Page’s advocate for this instalment because his submissions complete an argument his temporary literary companions never intended to construct.

There was no conspiracy, no secret committee and no mahogany-table discussion about suppressing the proletariat’s access to case law. Just three separate contributions, a technological revolution and an increasingly difficult question about who gets to understand the fucking law.

Angus Menzies And The Waterproofing Department

Angus Menzies has already provided TCAP with an unusually illuminating observation concerning Henderson v GCRM Ltd. The Employment Appeal Tribunal identified unanswered questions about possible manipulation of a dismissal decision. It returned the relevant issue to the original Tribunal.

Menzies anticipated “rather more water-tight findings to support its original decision”.

Our earlier article, The Outcome First, The Reasons Later, examined those words. We explored the uncomfortable relationship between a supposedly open question and an apparently anticipated conclusion. We need not excavate the same grave twice. What matters here is the broader implication: a judicial decision should follow a proper evaluation of evidence and law. Reasons should explain that journey. They should not function as weatherproof cladding installed after the desired conclusion has already moved into the fucking building.

His prediction does not establish that any judge intended to manipulate their reasoning. He may simply have believed the existing findings made the original outcome likely to survive reconsideration. The interesting question is what happens when outsiders can test that prediction. They no longer have to accept it merely as authoritative professional commentary.

The Specialist’s Disappearing Advantage

Previously, understanding why a judgment might contain a legal weakness could require specialist knowledge and expensive research facilities. It could also mean paying a solicitor to translate the whole affair into language resembling English. Today, a reader can search published judgments, compare statutory provisions and examine appellate authorities. Someone with little formal legal training can identify apparent gaps through sheer patience and access to the materials.

That does not make the amateur right. It makes the amateur considerably harder to exclude from the conversation. And it makes the quality of the underlying reasoning more important than the professional status of whoever happens to question it.

Daniel Rubin And The Social-Media Meat Grinder

Enter Daniel Rubin. His older commentary about Clarks’ gendered children’s shoes offered a pleasingly progressive celebration of public criticism. Rubin welcomed social media exposing everyday sexism. He warned that businesses perceived as failing to address discrimination could be “mauled across social media”.

TCAP has already examined the contrast between those sentiments and the treatment of my online criticism during the Cepac litigation. We do not need to revisit every exhibit, correspondence bundle or disputed publication. The original comparison remains available, and this argument is different.

Rubin discussed businesses, workplace equality and public reputation. He did not prescribe a disciplinary regime for judges. Nor did he suggest that Twitter storms should dictate judicial decisions. That extrapolation belongs to us.

But consider the principle beneath his commentary. Public information allows outsiders to identify conduct they consider objectionable and compare it with an institution’s stated values. They can then expose any discrepancy. Social media gives that criticism an audience beyond the institution’s control.

When The Spotlight Reaches The Bench

Now replace a company’s advertising promises with published judicial reasons. Replace the consumer examining a pair of shoes with a losing litigant examining an appellate authority. The question becomes whether an institution’s explanation withstands informed scrutiny when people outside its professional hierarchy can inspect the supporting material.

The legal system differs from a commercial organisation. Judges must remain independent of outside pressure. Quite right too. But independence from improper influence is not immunity from examination. Published judgments help the public scrutinise judicial reasoning. Criticism does not become inherently illegitimate merely because its author lost the case or dislikes the outcome. Nor must the critic hold a professional qualification permitting them to invoice somebody for noticing a contradiction.

Rubin recognised the uncomfortable consequences of public scrutiny when the subject was a shoe retailer. The modern question is what happens when the public starts reading the footnotes beneath the people doing the judging.

Mr B Gray And The ChatGPT Detector

Our rented contributor now enters from Page Outsourcing’s side of the courtroom. He brings a particularly interesting proposition about artificial intelligence.

The judgment in Thompson v Cepac and Page Outsourcing records Mr B Gray accepting that AI has legitimate uses. He nevertheless argued that my particular use had amplified abusive and obstructive correspondence. Paragraph 124 records his submission that AI-generated material gave rise to “no confidence that the Claimant understood what was being said”.

The Tribunal’s judgment reports counsel’s submissions; we do not have an independently verified verbatim transcript. Gray addressed my particular conduct and the reliability of future undertakings. He did not claim that every ChatGPT user was an intellectual vegetable. Those distinctions matter.

There is also a professional identification worth keeping precise. TCAP’s separate Justice Rangers investigation examined a possible connection between Page’s Mr B Gray and Benjamin Gray of Littleton Chambers. TCAP has not established the identification through a primary document. It remains provisional. We can examine the recorded submissions without converting an initial into a confirmed chambers affiliation.

When Assistance Becomes Suspicion

Nevertheless, the underlying inference deserves examination. A litigant uses software to research, structure and express arguments. The opposing advocate then questions whether the resulting correspondence reflects that litigant’s understanding. Assistance becomes part of the argument against confidence in the author.

That is a fascinating development from a profession which has long relied on clerks, pupils, junior lawyers and legal researchers. It also uses precedent banks and expensive subscription databases. Nobody ordinarily assumes that a barrister has forgotten the law because somebody else located an authority. Nor does a solicitor become incapable of understanding a document because a junior prepared its first draft.

Of course, AI introduces particular dangers. It can hallucinate authorities, misstate legislation and invent quotations. It can also produce polished nonsense with the confidence of a man selling timeshares on a sinking cruise liner. Those defects are real. Anyone relying on the output must check it. When an argument contains an error, identify that error, establish its significance and examine the responsibility of the person who submitted it.

Do not turn the mere appearance of artificial assistance into a substitute for evaluating the human being behind the words.

The Inconvenient Power Of A Search Box

Here lies the intellectual tension. When a litigant’s correspondence is objectionable, the author can bear responsibility for producing it. When the same litigant advances an AI-assisted argument, the opponent may question the extent of the author’s understanding. Those positions can coexist, but each requires evidence and proper distinctions. A fucking chatbot-shaped wand cannot establish either authorship or comprehension.

And technological assistance extends far beyond drafting emails. Someone using AI to organise a chronology can also ask it to locate statutory tests and relevant authorities. The same person can compare a judgment’s account of counsel’s submissions with the underlying evidence. They can then ask whether the findings actually support the result.

The machine is not the judge. It is not the litigant or a replacement for legal expertise. But it can provide extraordinarily useful reading glasses to somebody previously expected to admire the architecture from outside the fucking gates.

The Master Of The Rolls Has Entered The Chat

Unfortunately for anyone hoping to dismiss AI-assisted litigants as intellectually dependent passengers, one of the most senior judges in England and Wales has offered a rather inconvenient perspective.

On 18 August 2026, Sir Geoffrey Vos, the Master of the Rolls, delivered the John Lehane Memorial Lecture on artificial intelligence and the legal profession. He urged judges and lawyers to embrace the technology responsibly, improve access to justice and recognise how the public now uses digital tools.

Then came an especially interesting observation. Vos reported that judicial colleagues found many AI-assisted pleadings and skeleton arguments from litigants in person generally better than their pre-AI equivalents.

Well, fuck me. Perhaps the machinery has applications beyond producing correspondence whose author supposedly cannot understand it.

Vos did not propose accepting AI-generated arguments without examination. Quite the opposite. He insisted on human checking and responsibility for the final product. He also questioned applying different standards to AI when comparable mistakes by human researchers deserve the same criticism.

Competence Is A Question Of Evidence

That distinction gives the debate its proper intellectual weight. The question is not simply whether a litigant used AI. Ask whether the argument is accurate, whether the authorities exist and whether the reasoning holds up. Then ask whether the person submitting it can explain and substantiate it.

A polished paragraph may conceal misunderstanding. It may equally communicate a sound argument that the author would otherwise have struggled to express. The appearance of assistance cannot decide which possibility is true.

And if someone uses the same tools to examine a judgment, those tools do not determine whether the apparent discrepancy exists.

The legal profession has spent centuries distributing intellectual labour. Research, drafting, checking and advocacy have never been exclusively solitary activities. AI introduces new technical risks and questions of responsibility. It does not suddenly turn assisted reasoning into a species of fraud.

Somewhere in all this, the old assumption that legal understanding follows professional status is having a particularly unpleasant fucking afternoon.

The Judiciary Has Already Let The Robot Into The Building

The judiciary’s October 2026 position statement on AI makes the developing position clearer.

It describes a cautious but positive approach, including controlled pilots involving summarisation, transcription, chronology-building and document analysis. Human judicial officers remain responsible for reasoning, evaluating evidence and deciding cases.

Excellent. A perfectly sensible principle lies beneath the technology: assistance does not extinguish human responsibility.

The same distinction arose in Warrener-Iglesias v Coretech Security Services Ltd, an Employment Tribunal decision involving AI-assisted witness evidence. The judge distinguished ChatGPT’s generic material from the claimant’s own substantive account expressed with technological help. He compared some prompting functions to the way a solicitor might take instructions.

The decision did not endorse careless AI use. Nor did it establish the claimant’s substantive case. Its significance is narrower and more useful: help with expression does not automatically deprive an underlying human account of evidential value.

Open Justice Has Acquired Better Tools

Meanwhile, The National Archives offers a free official service through which people can examine published decisions. Its purposes include open justice, public understanding and scrutiny of legal reasoning. The database is not comprehensive, but the direction of travel is unmistakable.

A collection of servers and an enthusiastic bloke with broadband will not abolish the legal profession. Specialist expertise remains valuable and errors remain consequential. An AI-generated appeal can still be legally worthless, regardless of how impressive the headings look.

What is disappearing is the assumption that professional status alone provides an adequate answer to a substantive question.

A judge cannot dispose of an apparent inconsistency merely because the person who identified it used ChatGPT. Equally, a litigant cannot establish judicial misconduct simply by producing an AI-generated critique. Both arguments require evidence, applicable law and reasoning that withstands examination.

The fucking software changes access to the investigation, not the standard of proof.

The Difference Between Scrutiny And Intimidation

There is an important boundary. We should identify it before somebody drags this article into a courtroom attached to seventeen screenshots and a diagnosis of the author’s personality.

A judge must not decide a case according to which party has the larger social-media following or louder megaphone. Nor should the more unpleasant collection of supporters win. Threats, witness intimidation, improper disclosure and attempts to pressure decision-makers do not become legitimate accountability because somebody adds an attractive website template.

Equally, uncomfortable criticism is not automatically improper. Examining a published judgment, identifying a contested statement and comparing it with the evidence are ordinary exercises in public accountability. So is asking whether the Tribunal applied the correct legal test. An appeal remains the proper mechanism for correcting a legal error. Public commentary cannot reverse an order, but it can explain why that order deserves closer examination.

This distinction matters more as legal research grows accessible. An unrepresented litigant may misunderstand a judgment. So may a journalist, solicitor, barrister or AI system. To establish which interpretation holds up, examine the underlying material. Do not declare the debate closed merely because the wrong person has started reading it.

Public scrutiny does not replace judicial independence. It helps explain why judicial reasoning must withstand examination.

Three Contributions, One Very Uncomfortable Conclusion

Our temporary travelling circus finally comes together.

Angus Menzies predicted that an original decision would return with stronger waterproofing. Daniel Rubin explained that public exposure could leave institutions facing a social-media mauling. Mr B Gray questioned whether AI-assisted correspondence demonstrated sufficient understanding on its author’s part.

None of those statements proves corruption, collusion or even a shared philosophical position. Each argument has its own context, and TCAP has examined them separately.

The significance emerges when we set them against the modern accessibility of legal information.

The old professional hierarchy enjoyed an enormous practical advantage. Specialists knew where to find the law, how to interpret it and how to navigate procedural rules. Losing parties might disagree with the outcome without possessing the resources to discover whether their objections amounted to anything legally recognisable.

That advantage has not vanished. But it is eroding. A determined litigant can now build a chronology, research an unfamiliar doctrine and compare published authorities. They can also check quotations and seek independent criticism of the arguments. The resulting work may still be wrong. It can also expose a genuine error that an unassisted person might never have known how to articulate.

The Public Has Acquired The Building Plans

When that happens, the answer cannot simply be that the litigant should have left the thinking to somebody more expensive.

Judicial reasons do not belong intellectually to the judge who wrote them. Nor are they sacred texts whose meaning requires professionally approved interpreters. They explain exercises of public power. The public must be able to read, question and test them against the law.

The person challenging those reasons must take responsibility for accuracy. The people defending them should answer the substance. And the judiciary must continue distinguishing legitimate public examination from improper attempts to influence proceedings.

Fungus can bring the waterproofing. Rubian can issue the social-media weather warning. Benji can inspect the keyboard for traces of artificial intelligence. Although somebody should remind him that he’s only renting a room in this particular production.

The old gates are still standing. The lawyers still hold the keys to plenty of important doors. The judiciary still has its independence, and rightly so.

But the public has acquired the building plans.

And no amount of professional indignation is going to make the fucking foundations disappear.

Lee Thompson – Founder, The Cummins Accountability Project

Horsfield Menzies : Blog vs Behaviour – FUNGUS, RUBIAN & BENJI : THE JUDGMENT HAS NO FUCKING FIREWALL