Delegation, responsibility and the quiet rescue of the legal services sector: reflections on CILEX v Mazur

When Sheldon J handed down his decision in the High Court in September 2025, much of the legal services sector discovered, to its considerable alarm, that ordinary supervised litigation practice might have exposed firms and staff to criminal liability for nearly two decades. Paralegals issuing routine debt claims, law centre caseworkers assisting tenants facing eviction, personal injury clerks pressing “submit” on the Money Claims Online portal – all were, on the High Court’s reading of the Legal Services Act 2007, potentially exposed to criminal liability under section 14, subject to the Act’s defence for those who did not know, and could not reasonably have been expected to know, that an offence was being committed. There was also the added indignity of possible contempt of court under section 14(4).

The Court of Appeal’s decision in March 2026, given principally by Sir Colin Birss, Chancellor of the High Court (Birss C), draws that conclusion back from the cliff edge. But it does so in a way that repays careful reading, because the reasoning is more interesting – and more fragile – than the comfortable headline result might suggest.

This article considers what the Court of Appeal actually decided, why the reasoning matters beyond the immediate scramble for regulatory cover, and what the judgment tells us about the deeper structural question lurking in the 2007 Act: who, in a modern law firm or law centre, is really “practising law”?

The problem in miniature: Mr Middleton and the Money Claims Online form

The facts are almost comically mundane. Charles Russell Speechlys instructed Goldsmith Bowers to recover unpaid fees from Mrs Mazur and Mr Stuart. A Senior Litigation Executive at GBS, Mr Middleton, drafted the claim, submitted it through Money Claims Online, and signed the Particulars of Claim as “Head of Commercial Litigation”. Mr Middleton had been suspended as a solicitor in 2008 and held no practising certificate. The SRA had, however, given GBS permission to employ him as a Senior Litigation Executive. That permission had originally required supervision by Mr Bowers, but the most recent approval required Mr Middleton’s work to be directly supervised by Mr Ashall.

Mrs Mazur and Mr Stuart, litigants in person throughout, spotted this and applied to strike out the claim. What followed is a case study in regulatory confusion. The SRA initially wrote on 2 December 2024 that GBS’s employees “are permitted to undertake ‘reserved activities’ due to Section 21(3)” of the 2007 Act.

Section 21(3) states:

In this section “regulated persons”, in relation to a body, means any class of persons which consists of or includes—

(a) persons who are authorised by the body to carry on an activity which is a reserved legal activity;

(b) persons who are not so authorised, but are employees of a person who is so authorised.

By the time the matter reached Sheldon J, the SRA had completely reversed course. The judge put three questions to the Law Society and SRA, asking whether a non-admitted person employed by an SRA-regulated firm was:

• permitted to support an authorised solicitor in conducting litigation;
• permitted to conduct litigation under the supervision of an authorised solicitor; and
• permitted, by virtue of the firm’s authorisation, to conduct litigation themselves as an employee of the regulated entity.

Both bodies answered “yes”, “no” and “no”. Supporting was lawful; conducting under supervision was not; and the firm’s authorisation did not cover the employee. Sheldon J, understandably, accepted the aligned submissions of the SRA and the Law Society.

The regulators’ about-face is, in retrospect, the key dramatic moment. As Birrs C pointedly observes at [196], “both DDJ Campbell and HHJ Simpkiss were faced with shifting sands and arguments that were wrong and later abandoned.” The sense that the judgment below rested on regulatory advice the regulators themselves could not consistently defend hangs over the whole appeal.

The central question: what does “carry on the conduct of litigation” mean?

The Court of Appeal reduced the case to a single question of statutory construction. Section 14(1) makes it an offence “for a person to carry on an activity … which is a reserved legal activity” without entitlement. Section 12(1)(b) lists “the conduct of litigation” as such an activity. The meaning of “carry on the conduct of litigation” therefore does all the work.

Three rival constructions were before the court:

1. The LSB’s position: “carry on” simply means “do”. The offence catches anyone who performs tasks within the conduct of litigation, regardless of whether they do so for an authorised person.
2. The Law Society/SRA position (as it evolved): an authorised person must “direct and control the performance of the activity”. Unauthorised people can “assist” or “support” but not conduct litigation under supervision. Initially this required “universal prior approval” of every document. As the court noted at [11], that was a remarkable submission, because it highlighted that many solicitors may have been operating on a model which, if the submission were right, exposed them to criminal liability whenever they allowed an unauthorised person to issue proceedings without specific prior approval.
3. CILEX’s position: “carry on the conduct of litigation” has two dimensions – performance and responsibility. Performance can be delegated; responsibility cannot. An authorised individual who retains responsibility is the one “carrying on” the activity, even if an unauthorised employee actually does the task.

The court unanimously adopted the third view. The words “conduct of litigation” refer to the tasks; the words “carry on” refer to “direction and control of, and responsibility for, those tasks” ([21]).

Birrs C gives six reasons at [163] – [169]:

• The legislative meaning of “carry on” has not changed significantly since Victorian solicitors legislation; it has a long statutory pedigree connoting more than mere doing.
• While “carry on” applies to all six reserved activities, that does not force a single meaning; different activities engage different considerations.
• “Conduct” alone cannot bear the weight of meaning “direct and control”; the schedule 2 definition makes clear “conduct of litigation” refers to the tasks themselves.
• Graham v Lewis (1888) 22 QBD 1 supports reading “carry on” as connoting management and control – a clerk employed by a solicitor was not “carrying on business” in the City; the solicitor was.
• Paragraph 1(7) of schedule 3 (the rights of audience exemption for assistants) does not compel the contrary inference; different reserved activities have their own histories.
• Sections 15 and 16 of the 2007 Act are not engaged because the unauthorised employee is not, on the court’s analysis, “carrying on” the activity at all.

Why the historical analysis matters

The most intellectually substantial part of the judgment is the long historical section at [122] – [148], tracing delegation from The Law Society v Waterlow Bros & Layton (1883) 8 App Cas 407 (HL) through Myers v Elman [1940] AC 282 and the “1946 Managing Clerks’ Gazette” article* to Hollins v Russell [2003] EWCA Civ 718 and Agassi v Robinson (Inspector of Taxes) (Bar Council intervening) [2005] EWCA Civ 1507. This is not just scene-setting: it does the crucial work of establishing what Parliament must be taken to have known when it enacted the 2007 Act.

Waterlow is the linchpin. The Earl of Selborne LC accepted that the statutes requiring acts to be done by qualified solicitors “did not require the solicitor to do those things personally. The acts could be done in the solicitor’s name by a competent agent and in such a case the agent acting in that way did not incur the penalties”. The quoted passage – “The existence of a real solicitor in every one of these cases, really applying for probate, initiating the whole matter, carrying on the whole business in his own name, although not personally going to the office, is a fundamental fact in the case” – is essentially the CILEX position transported back 143 years.

Lord Atkin’s observation in Myers v Elman is equally striking: “Such delegation is inevitable … The machinery of justice would not work without them.” When this statement was made in 1940, no judge thought it controversial that an unqualified managing clerk might, in practice, run a solicitor’s litigation caseload. The 1946 Managing Clerks’ article, which both CILEX and the Law Society tried to claim, is read by Birss C as plainly contemplating that “a managing clerk can and would then, in a proper case, take steps and make decisions without the prior approval of their principal” ([133]).

The inference the court draws is compelling: Parliament in 2007 must be taken to have known of this practice, and nothing in the legislative materials or the Clementi Review suggests an intention to criminalise it. The 2007 Act was liberalising – but, as Birrs C carefully notes at [150], the liberalisation was “in relation to the diversity of regulatory bodies, not the offences”. This is an important distinction. The Act fundamentally rearranged who regulates the profession; it did not, on the court’s reading, fundamentally rearrange what the profession may do.

The distinction that never was: “assisting” versus “conducting under supervision”

Sheldon J, following the Law Society and SRA, had drawn a line between (a) “supporting (or assisting) an authorised solicitor in conducting litigation” and (b) “conducting litigation under the supervision of an authorised solicitor”. The first was lawful; the second was not. The Court of Appeal demolishes this distinction, and the demolition reveals something important about the earlier reasoning.

The distinction was always linguistically strained. “Support” and “assist”, as the Law Society itself conceded, “mean much the same thing” ([9]). The practical difference between an assistant who prepares a claim form for a solicitor to check and an unauthorised person who prepares and files it under supervisory arrangements is not obvious. And as the case progressed, the Law Society’s insistence on “universal prior approval” as the supposed hallmark of (a) collapsed. First the SRA introduced a qualification for “rigorous controls and systems” allowing unauthorised staff to issue simple debt claims without prior sight by an authorised individual ([12]). The Law Society adopted it. At that point, as CILEX argued in reply at [115], the distinction was intellectually finished:

“The only logical legal justification for this not being an example of ‘carrying on the conduct of litigation’ by the unauthorised person must be because the authorised individual remains responsible for that act even though they have not approved it in advance.”

Once the regulators accepted that unauthorised staff could lawfully do the thing without prior approval, the only coherent explanation was that the authorised individual’s ongoing responsibility was doing the work of keeping the activity within the law. That is exactly the two-dimensional construction CILEX had been pressing all along.

The schedule 3 argument: a silence that says nothing

One of the strongest textual arguments against CILEX came from the structure of schedule 3 itself. Paragraphs 3 and 4 provide express exemptions for unauthorised persons carrying out reserved instrument activities and probate activities “at the direction and under the supervision” of an authorised individual. Paragraph 2, which deals with the conduct of litigation, contains no such exemption. The expressio unius inference is obvious: Parliament knew how to create a supervision-based exemption and chose not to for litigation.

Birrs C’s answer at [155] is honest but uncomfortable:

“one cannot safely read across from the detailed drafting of provisions about one reserved legal activity to make inferences about what was intended in respect of another reserved legal activity. They are separate activities with their own histories and practices.”

This is plausible, but it is not a knockout. The argument really requires the court’s other reasoning – particularly the historical analysis and the two-dimensional reading of “carry on” – to carry the weight. The schedule 3 point is defused rather than defeated.

There is a second, more elegant response buried in the construction: if the CILEX reading is right, no exemption is needed for litigation because the unauthorised person is not, as a matter of construction, “carrying on” the activity in the first place. The exemptions in paragraphs 3 and 4 exist because reserved instrument and probate activities require a different drafting approach, not because litigation is somehow uniquely non-delegable.

What Ndole and Baxter were really about

The judgment’s second doctrinal achievement is the rehabilitation of Baxter v Doble [2023] EWHC 486 (KB) (Baxter). In the months after Sheldon J’s judgment, Baxter was widely read as authority for an expanded list of tasks constituting the conduct of litigation – a reading that threatened to pull things like witness statement preparation squarely within the scope of section 14.

Birrs C’s careful dissection at [182] – [186] explains what Baxter actually did. Both Ndole Assets v Designer M&E Services [2018] EWCA Civ 2865 (Ndole) and Baxter involved unauthorised people working for litigants in person. In that situation there is no authorised individual at all – and litigants in person have, by Gregory v Turner [2003] EWCA Civ 183, a personal non-delegable right. The question in those cases was therefore a binary one: had the unauthorised person, by taking responsibility for the tasks, themselves crossed the line into carrying on the conduct of litigation? The “fact and degree” test from Ndole at [67] – [68] was directed to answering that question, not to defining the boundary of the reserved activity itself.

As Birrs C puts it at [178]:

“Ndole decided that the question of whether an unauthorised person acting for a litigant in person was carrying on the conduct of litigation was a matter of fact and degree. The question was whether the task was purely mechanical or whether the unauthorised person had assumed responsibility for it. That, as I have said, is not the question here.”

This matters enormously for practice. An unauthorised person working within a regulated firm or law centre for and on behalf of an authorised individual is in a fundamentally different position from an unauthorised “legal consultant” working for a lay client. The former is within a responsibility structure; the latter is the responsibility structure. The court’s refusal to collapse these two situations into a single test is, on reflection, the most consequential move in the judgment.

The deeper question: who practises law?

Stepping back, Mazur forces a question the 2007 Act has never really answered: what does it mean to practise law in a modern legal services market?

The traditional answer, embedded in Victorian legislation, was individual: the solicitor was the practitioner, personally responsible, personally liable, personally present. The Waterlow reasoning was already a concession to reality – solicitors could not physically be in every probate registry, so the agent doctrine allowed the principal to be present through another. But the principal’s identity and responsibility remained central.

The 2007 Act complicates this. It authorises entities (law firms) as well as individuals. It creates regulated persons (section 21(3)) who are not authorised persons. It operates against a background in which a large London firm might employ hundreds of paralegals, trainees, legal executives and other staff, many of whom routinely perform work that looks indistinguishable from what a junior solicitor might do. The traditional model – in which the solicitor on the record is personally doing, or personally approving, everything that happens – is not how the sector has functioned for decades, if it ever did.

The Court of Appeal’s answer is essentially pragmatic: “practising law”, in the sense of “carrying on the conduct of litigation”, is about responsibility structures rather than individual task performance. As [187] puts it:

“An unauthorised person may lawfully perform any tasks, which are within the scope of the conduct of litigation, for and on behalf of an authorised individual … provided the authorised individual retains responsibility for the tasks delegated to the unauthorised person (both formal responsibility and the responsibilities identified at section 1(3) of the 2007 Act).”

The reference to section 1(3) is important and, I think, under-noticed. The “responsibility” the authorised individual must retain is not just formal or legal. It includes the professional principles: independence and integrity, proper standards of work, acting in the client’s best interests, duty to the court, and client confidentiality. This is a substantive obligation, and it does real work in limiting the judgment’s reach.

Lady Justice Andrews’s short concurrence at [198] makes the point crisply:

“In essence, the question in any given set of circumstances will be whether the unauthorised person, in carrying out whatever tasks which fall within the scope of ‘conduct of litigation’ have been delegated to him or her, is in truth acting on behalf of the authorised individual. If they are, it is the authorised individual who is conducting the litigation. But if the reality is that the litigation is not being conducted by the unauthorised person for and on behalf of the authorised individual, they will be committing an offence.”

The “in truth” does a great deal of work in that passage. A merely nominal supervisor who is not supervising in reality – Birrs C’s example at [116] of an authorised individual on a year’s sabbatical – cannot sensibly be said to be carrying on the litigation. The protection of the regulatory regime is not secured by paper supervision.

What the judgment does not decide

The case is, however, narrower than its “real world impact” suggests, and the judgment is admirably candid about what it leaves open.

First, the court declines to attempt a comprehensive list of tasks within the conduct of litigation. At [29]: “it is simply not possible to provide a comprehensive list of all those tasks that fall within and outside the conduct of litigation.” Seven items are identified at [193] as “unlikely” to fall within the definition – pre-litigation work, giving legal advice, correspondence, gathering evidence, instructing experts, signing statements of truth, and signing documents the CPR permits a legal representative to sign. But the court was careful to say that these items were neither challenged nor debated, and should be regarded as common ground rather than the product of adversarial argument. The scope of the second limb of paragraph 4(1) (“commencement, prosecution and defence”) remains contested, with the authorities divided on whether preparing a witness statement falls within it.

Second, the judgment does not address the position where there is no authorised individual at all – the Khan, Agassi and Baxter line. That framework remains intact. The unauthorised “legal consultant” acting for a lay client is still on dangerous ground, and the Ndole fact-and-degree test still applies to them.

Third, the scope of section 23 (transitional protection for non-commercial bodies) and the correctness of SRA v Khan [2021] EWHC 3765 (Ch) are explicitly left undecided at [194]. The law centres get the answer they want by the front door and need not rely on these fallback arguments.

Fourth, and perhaps most significantly, the judgment does not prescribe what supervision must look like. At [187]: “The delegation of tasks by the authorised individual to the unauthorised person requires proper direction, management supervision and control, the details of which are a matter for the regulators.” This is an invitation – almost an instruction – to the SRA, CILEX Regulation, the Bar Standards Board and others to clarify their guidance. The LSB’s interim report of 29 January 2026, quoted at [89], already acknowledges that guidance “differed across parts of the regulated sector”. The period ahead will be one of regulatory rewriting.

A note on regulatory confusion

No analysis of Mazur should pass over the observations about regulatory performance. The SRA’s letter of 2 December 2024, on which HHJ Simpkiss partly relied, was, in the Court of Appeal’s words, “wrong in law and contrary to the submission made to DDJ Campbell” ([196]). The SRA then reversed position. Mr Stuart’s submission at [114] – that “the real problem here was a gross failure of regulation” – is not adopted by the court, but the judgment does not contradict it either. There is something uncomfortable about a criminal regulatory regime in which the principal regulator can, within the space of a few months, tell one court that employees “are permitted to undertake ‘reserved activities” and then adopt the opposite position before another.

Mrs Mazur’s plainly stated complaint at [114] – that “the 2007 Act does make fairly clear who can conduct litigation but does not explain to anyone what the conduct of litigation is” – is also not really answered. A statute that creates a two-year imprisonment offence turning on the difference between (a) preparing a claim form for approval and (b) preparing and submitting it under general supervisory arrangements is a statute that citizens cannot realistically read and understand. That is not the court’s fault, but the judgment’s elaborate reconstruction of what Parliament must have meant – building on Victorian case law, 1946 trade journal articles, and inferences from structural silence – rather proves the point.

Conclusion: a workable settlement, but not a tidy one

The Court of Appeal has reached the right result. An alternative outcome would have meant either retrospectively criminalising a vast swathe of ordinary legal practice, or requiring the legal services sector to reorganise itself into a shape it has not taken for well over a century. Neither would have been a sensible reading of the 2007 Act against its context.

But the reasoning gets there by a route that is more patchwork than principled. The court relies on:

• A long historical practice that does not map perfectly onto the 2007 Act’s modern entity-based framework
• A two-dimensional reading of “carry on” that the LSB credibly argued against
• A treatment of the schedule 3 structural argument that is plausible but not conclusive
• A rehabilitation of Baxter that requires reading that decision more narrowly than the reported headnote suggests
• An appeal to section 1(3) professional principles that gives substantive content to “responsibility” in a way that is not immediately obvious from the statutory text

The result is a legal framework that works – unauthorised staff can continue doing the work they have always done, provided their authorised principals genuinely supervise – but which depends heavily on the good faith of regulators in defining what genuine supervision looks like. The SRA and CILEX Regulation now bear the weight the 2007 Act arguably should have borne itself.

Perhaps the deepest lesson of Mazur is that “practising law” and “being entitled to practise law” have come apart in a way the 2007 Act never quite confronts. The regulated individual is the bearer of responsibility; the unregulated employee is often the doer of tasks. The court has accepted that this is a coherent state of affairs and has given it statutory footing. Whether Parliament, if asked, would recognise its handiwork is a different question. The “fundamental fact” the Earl of Selborne identified in 1883 – the existence of a real solicitor, really carrying on the business – remains the touchstone. What has changed is everything around it.

* ‘The Duties and Responsibilities of a Managing Clerk’ (June 1946) The Solicitors’ Managing Clerks’ Gazette 29.