Sir Robert Buckland delivered a keynote speech at the Costs Law Reports Annual Conference held at Lincoln's Inn on 24th September 2026. Speaking about the tsunami of AI "slop" hitting the courts at present, he looked at the existing costs provisions and called for greater use to be made of the provisions relating to unreasonable conduct when it comes to the huge and increasing amount of time having to be expended on examining AI submissions that are often inaccurate, verbose and unhelpful when it comes to the swift administration of justice. His full speech is here:
The Price of Justice: Civil Costs, Litigation Funding and the AI Problem.
Address to the Costs Lawyers Conference, 24/09/2026
Rt Hon Sir Robert Buckland KBE KC
It’s fair to say that in the order of political and policy priorities, civil justice and costs do not feature at the top of the list. At the Ministry of Justice, the focus, frankly, is mainly on the criminal justice system, both in terms of the systemic issues that it poses and the day-to-day high-profile challenges that it can create. Jack Straw, one of my predecessors, put it to me once that as Lord Chancellor and Justice Secretary with responsibility for, amongst other things, prisons and prisoners, you are never more than a few hours away from a real crisis.
I was a criminal practitioner for twenty years or so prior to my entry into frontline politics, but my work as a government law officer and then at the MOJ has meant that I have diversified.
I think costs are, actually, important. The cost of litigation is not simply the price of legal work.
It is also a consequence of human behaviour.
And that, I think, is the point at which the modern debate about civil costs needs to begin.
We have spent the last fifteen or twenty years trying to make civil litigation more proportionate.
We have had Jackson.
We have had costs budgeting.
We have had the refinement of proportionality.
We have had fixed recoverable costs.
We have had guideline hourly rates.
We have had reforms to litigation funding.
And now we have artificial intelligence.
Each development has been presented, in one way or another, as a means of making justice more efficient.
But there is a danger in confusing the cost of producing legal material with the cost of administering justice.
AI makes that distinction particularly stark.
A machine may produce a fifty-page document in seconds.
That does not mean that the fifty pages cost nothing.
Someone has to read them.
Someone has to check the authorities.
Someone has to answer them.
Someone has to decide which parts matter.
And, ultimately, a judge has to work out what the case is actually about.
The production cost may have collapsed but the cost to the justice system has increased.
That is the central issue I want to explore.
And it leads to a wider proposition:
The costs jurisdiction should not merely allocate the bill after litigation has happened. It should be used to influence the behaviour which creates the bill in the first place.
The civil costs revolution
The modern civil costs regime is built around a perfectly defensible principle:
proportionality.
A £50,000 dispute should not ordinarily generate £500,000 of costs.
But proportionality is not the same as cheapness.
That distinction matters.
A case can be cheap because it has been efficiently conducted.
It can also be cheap because necessary work has simply not been done.
Those are very different things.
The challenge for the costs system is therefore to distinguish between waste and necessary expenditure.
The expansion of fixed recoverable costs has been an important part of the response.
It brings greater predictability.
It makes litigation economics easier to understand.
And it can prevent the costs tail from wagging the substantive dog.
But fixed costs inevitably involve a degree of approximation.
A case which looks like a £50,000 case from the outside may contain an unusually difficult legal issue, an enormous disclosure exercise or a technical evidential problem.
The costs regime therefore has to retain mechanisms for exceptional cases and unreasonable conduct.
And that is why the conduct provisions of CPR 44 are so relevant.
They remind us that costs are not merely about value.
They are about behaviour.
The court can consider whether it was reasonable to raise, pursue or contest an allegation or issue, and the manner in which a party has conducted the proceedings.
That is not a marginal power.
It is one of the principal mechanisms through which the court can tell litigants:
You are entitled to litigate. You are not entitled to litigate irresponsibly.
Costs are also relevant to access to justice. That takes us naturally to litigation funding. I declare an interest as a Class Representative in an ongoing claim in the CAT relating to alleged anti-competitive behaviour by some crypto currency companies that caused loss to consumers. I don’t talk about the details of that case, but instead focus on current policy developments.
The Supreme Court's decision in PACCAR transformed the funding landscape.
The essential issue was the statutory definition of a damages-based agreement.
The Supreme Court concluded that certain litigation funding agreements in which the funder's remuneration was calculated by reference to damages recovered could fall within that definition.
That was consequential because damages-based agreements are subject to a statutory regime which many conventional third-party litigation funding agreements had not been structured to satisfy.
The effect was to place the enforceability of significant numbers of funding arrangements in doubt.
And that mattered particularly in collective proceedings, which are often economically impossible without external capital.
A consumer with a claim worth £100 may have no realistic prospect of financing litigation costing millions.
But if ten thousand consumers have suffered the same £100 loss, the aggregate claim may be enormous.
That creates a classic access-to-justice problem.
The Government has accepted that PACCAR needs to be reversed.
The Civil Justice Council's Litigation Funding Working Party reported in June 2025. It recommended proportionate regulation of third-party funding and recognised its role in access to justice.
The Government has accepted the two central recommendations:
First, legislation to mitigate the effect of PACCAR.
Secondly, proportionate regulation of litigation funding agreements.
The position as at September 2026 is therefore quite clear in policy terms, but incomplete in legislative terms.
The Government remains committed to legislation.
But the legislation has not yet been enacted.
Indeed, on 3 September Lord Lemos told the House of Lords that the Government was urgently seeking a legislative vehicle. He said the legislation would clarify that litigation funding agreements are not damages-based agreements and would thereby mitigate the effect of PACCAR.
That delay is not without consequences.
Every month in which the position remains uncertain is a month in which funders, claimant lawyers and defendants have to price uncertainty into litigation.
And uncertainty itself has a cost.
The previous Government attempted to address the issue through the Litigation Funding Agreements (Enforceability) Bill in 2024.
That Bill sought to reverse the effect of PACCAR retrospectively.
The present Government has not adopted that approach.
It has indicated that the eventual legislative solution will operate prospectively.
That distinction creates difficult transitional questions.
What happens to agreements entered into before the new legislation?
What happens to litigation which has been restructured because of PACCAR?
What happens to settlements negotiated on the assumption that an existing LFA might be unenforceable?
And, perhaps most importantly, how do we avoid replacing one period of uncertainty with another?
These are precisely the sorts of questions which create satellite litigation.
And satellite litigation is the enemy of proportionate costs.
But I think the Government is right about one thing which is sometimes missed in this debate.
Reversing PACCAR does not have to mean returning to an entirely unregulated funding market.
Indeed, the CJC's recommendations point in the opposite direction.
The objective should be proportionate regulation.
That means asking some rather practical questions.
What information should a funded claimant receive?
What control should a funder have over litigation strategy?
What happens when the interests of the funder and claimant diverge?
Should funders' returns be capped in some circumstances?
What financial resources should a funder be required to maintain?
What disclosure should be made to the court?
And how should funding interact with security for costs?
These are not anti-funder questions, Nor are they pro-funder questions, but they are questions about market architecture.
The Government has expressly recognised that litigation funding needs to remain commercially viable for funders while also being fair and transparent.
That is the correct balance.
My friend Simon Davis will talk about the Access to Justice Foundation and I won’t trespass on his bailiwick save to say in advance “What he said”!
But I want to turn to the development which I suspect will occupy the courts much more in the next five years than PACCAR.
Artificial intelligence.
We need to be sensible about this.
AI is not the enemy.
Used properly, it is an extraordinarily useful tool.
It can help a litigant understand a procedural rule.
It can assist with research.
It can identify patterns in a large body of documents.
It can help a lawyer organise material.
It may make some legal services cheaper.
All of that is potentially very valuable.
But there is a fundamental distinction between using AI to assist judgment and using AI instead of exercising judgment.
The second is dangerous.
And we are already seeing the consequences.
There have been cases in which lawyers have filed authorities which did not exist or which did not say what they were said to say.
That is not a theoretical risk.
And it is precisely why the judiciary has been taking this issue seriously. Lord Briggs recently described the AI wave as a “tsunami” and he’s not wrong.
The Civil Justice Council's AI working group reported in June 2026 that there is currently a strong consensus that additional formal AI-specific requirements are not necessary for professional legal drafting: existing professional responsibility frameworks should generally be sufficient.
But the CJC has identified particular issues concerning witness statements, expert evidence and litigants in person, and its final report is expected later this year.
I think that is sensible.
We do not need a new rule for every new piece of software.
But we do need a clear statement of responsibility.
If you put it before the court, you own it.
The fact that ChatGPT, Claude, Gemini or some other system generated it is not a defence.
This brings us to what we have come to know as AI slop.
It is not a criticism of AI itself.
It is a description of a particular way of using it.
The production of large quantities of superficially impressive but unnecessary legal material.
A litigant can now take a grievance and ask an AI system to turn it into a pleading.
Then a witness statement.
Then a skeleton argument.
Then a list of authorities.
Then submissions in reply.
Then a response to the reply.
The machine does not become tired.
The litigant does not become embarrassed by the size of the document.
And, unlike a human lawyer, the machine has no instinctive sense that a particular argument is not worth making.
That creates a new costs problem.
The person generating the material may incur virtually no marginal cost.
The opponent does.
The court does.
The taxpayer does.
And the judicial system does.
This is potentially a profound change in the economics of litigation.
Historically, prolixity had a natural brake.
Someone had to pay a lawyer to write it.
AI removes that brake.
And that means that the costs jurisdiction needs to provide another one.
I do not think we need an “AI costs rule”.
We already have the machinery.
CPR 44 contains the necessary principles.
he answer is therefore to make clear that AI-generated conduct is still conduct.
If I file a document containing ten fabricated authorities, it does not become less unreasonable because the hallucination was committed by a machine.
The question remains:
Was it reasonable to do this?
That is the right question.
And what about litigants in person?
Here I would introduce an important qualification.
We must not treat litigants in person as second-class litigants.
Many are unrepresented because they cannot afford legal advice.
Some have no legal aid entitlement.
Some have been unable to find a lawyer.
And some are involved in cases in which representation is simply impractical.
The answer therefore cannot be:
“Litigant in person plus AI equals costs sanction.”
That would be both unfair and unworkable.
There needs to be a distinction between uninformed use and irresponsible use.
A person who genuinely believes an AI-generated proposition to be correct may need education.
A person who is told that an authority does not exist and continues to rely upon it may be behaving very differently.
Similarly, a litigant who produces a long document because they genuinely cannot understand what the court requires is not in the same position as one who deliberately floods the court with irrelevant material after being warned to stop.
The court needs discretion.
But discretion requires the willingness to use it.
I would favour a relatively simple development.
At the outset of litigation, particularly where a litigant in person is involved, the court could make clear that:
- AI may be used as an aid;
- responsibility for accuracy remains with the person filing the document;
- authorities must be verified;
- factual assertions must be checked;
- documents must be proportionate;
- irrelevant or repetitive material should not be filed;
- and unreasonable costs generated by the use of AI may be taken into account when costs are determined.
That would achieve something important.
It would put litigants on notice.
And it would give the costs judge a principled basis for acting later.
The issue becomes particularly interesting in the fixed-costs regime.
Fixed costs can create a difficulty because the innocent party may have incurred significant work dealing with unreasonable conduct but may not recover all of that work through the normal fixed-cost mechanism.
That is why the existing provisions concerning unreasonable behaviour are so important.
The system should be capable of recognising the difference between:
“This case was difficult.”
and
“This case became expensive because one party behaved unreasonably.”
If the second is true, there should be consequences.
Otherwise fixed costs can create an unfortunate incentive.
A party may calculate:
“The other side has to deal with everything I produce, but their recovery is fixed anyway.”
That would be precisely the wrong result.
Costs rules should not reward procedural aggression.
The position of professional lawyers is, of course, different.
A litigant in person may reasonably say:
“I did not know.”
A regulated professional has less room for that answer.
A solicitor or barrister is expected to exercise judgment.
If a lawyer signs a pleading, skeleton argument or witness statement, professional responsibility remains with the lawyer.
The CJC's current position—that existing professional responsibility frameworks are broadly sufficient for professional AI use—is therefore important.
But “existing professional responsibility” only works if it is enforced.
And enforcement does not necessarily mean disciplinary proceedings.
Costs can sometimes provide a more immediate and proportionate response.
There is a useful analogy in the employment tribunal system.
Its costs jurisdiction is different from the ordinary civil courts.
But it recognises a principle which is worth preserving:
unreasonable conduct can have financial consequences.
That is important because the costs jurisdiction is not supposed to be purely compensatory.
It also has a behavioural function.
It tells litigants:
You are entitled to pursue your case.
You are entitled to be wrong.
You are entitled to lose.
But you are not entitled to behave unreasonably without consequence.
There is another piece of reform which deserves to be brought into the same conversation.
The distinction between contentious and non-contentious business increasingly looks like a historical artefact rather than a sensible organising principle for costs regulation.
The Johnson Committee's work therefore has importance beyond solicitor-client bills.
It is part of a broader question:
what should a modern costs system actually look like?
There is, however, a warning here.
If we abolish an outdated costs procedure, we must not simply move the work somewhere else.
If a £10,000 solicitor-client costs dispute which would once have gone to the High Court is instead transferred to the Legal Ombudsman, the Ombudsman needs the resources and jurisdiction to deal with it properly.
Otherwise we have not solved the problem.
We have simply moved the queue.
Good reform should reduce the total cost of the system.
There is a phrase I have used in another context which seems appropriate here:
a language of responsibility rather than power.
Everyone involved in civil justice has responsibilities.
Judges have responsibility for proportionate case management.
Lawyers have responsibility for the advice they give and the material they put before the court.
Funders have responsibility towards the claimants whose litigation they finance.
Litigants have responsibility for the cases they bring.
And Government has responsibility for ensuring that the legal system has the resources necessary to function.
But there is another responsibility which is increasingly important.
If you introduce a new technology into litigation, you also accept responsibility for the costs that technology creates.
Unreasonable conduct has consequences.
Costs caused unnecessarily should, so far as justice permits, fall upon the person who caused them.
And access to justice requires us to distinguish between the cost necessary to vindicate a right and the cost created by the way in which somebody chooses to litigate.
That, I think, is where the modern civil costs debate should now go.
We should aim to end up with having something much more valuable:
a civil justice system in which the resources of the court are devoted to deciding real disputes, rather than managing the consequences of the way in which people choose to conduct them.
And next time you are in the pub and your fellow drinker asks what sort of lawyer you are, you can stifle their yawns by referring them to my address to you today and reminding them that costs are, actually, important!
