reportAI Governance

Legal AI by the numbers: thirteen surveys, five vendors and the governance gap, with the methods exposed

Every 2024 to 2026 survey of AI in legal practice on one schema: adoption runs 30% to 92% depending on who asked; 43% of firms have no policy, 54% no training.

Lexis+ AI

Edited and verified by Cognesio LLP · updated

Researched with AI assistance · sources verified by Cognesio LLP · How this was made ↓

Adoption of AI in legal work is high wherever a survey has asked, in the United States, the United Kingdom, Europe and China; the governance around it is not keeping pace. That was the finding of the first edition of this page in July 2026, resting on four surveys, and it holds in this second edition, which rests on thirteen. What the thirteen also show is that the headline adoption number depends almost entirely on who asked the question and how: 30 percent of American private-practice lawyers in one 2024 survey, 92 percent of lawyers in eleven countries in another fielded a year later, 69 percent of legal professionals in a third, and 15 percent “dependent” in a fourth. The numbers are not contradictory. They measure different things, for different sponsors, in different words.

This report puts every survey of AI in legal practice published between September 2024 and September 2026 that the site has been able to read on one schema: who paid, who was asked, how many, when, what the question said, and what the number is. It then triangulates the three governance figures that survive the comparison: fewer than half of firms have any AI policy, about one in ten has one it enforces, and roughly half have provided no training. It does not survey anyone; it is a register of surveys, with the methods exposed, so that a reader can see which numbers can be compared and which cannot.

Key findings

  1. Thirteen surveys published between September 2024 and 2 September 2026 report AI use or governance in legal practice. Eight are vendor-sponsored, from five vendors (LexisNexis twice, Thomson Reuters three times, Wolters Kluwer, Clio, 8am), three come from professional bodies (the ABA, ILTA, the IBA with CAIDP), one from an industry association (CLOC, on Harbor’s data) and one from an intergovernmental body (UNESCO). Sample sizes run from 116 to 2,527; two surveys publish no sample size at all, and a third publishes one only for part of its sample.
  2. “Adoption” is measured at least five ways: personal use of any AI tool (Wolters Kluwer 92 percent; 8am 69 percent; LexisNexis UK 61 percent), organisation-wide use (Thomson Reuters 40 percent), use of AI “in practice” by private-practice lawyers (ABA 30 percent), firms using or exploring generative AI (ILTA 80 percent), and dependence (LexisNexis UK 15 percent of large-firm lawyers). No two of those are the same quantity.
  3. Firm size explains more variance than year. The ABA’s 2024 survey found 46 percent of firms with 100 or more lawyers using AI against 18 percent of solos; ILTA’s 2025 survey found 100 percent of firms with 700 or more lawyers using or exploring generative AI against 63 percent of firms under 50; Clio’s 2026 solo and small-firm survey found 71 and 75 percent.
  4. The policy figure is the most consistent across sponsors: the IBA found 43 percent of 210 firms had any AI policy in 2024; ILTA found 19 percent of 580 firms had no policy in 2025 (down from 34 percent) and 45 percent permitted use of vetted or sanctioned tools and applications (a multiple-response item, not an exclusivity rule); 8am found 43 percent of respondents’ firms had no policy and no plans for one, and 9 percent had a written policy that was enforced, in late 2025.
  5. The training figure is the widest gap in the record: UNESCO found 9 percent of judicial operators had received any AI training against 44 percent who had used AI (2024); 8am found 54 percent of firms provided no training and had no plans to, with 11 percent mandating it (late 2025); Thomson Reuters’ 2026 Future of Professionals survey found 43 percent of professionals naming untrained people as the reason AI strategy was not delivering.
  6. The 2026 vendor surveys have moved from adoption to disappointment: Thomson Reuters reports 91 percent of 1,816 professionals saying their organisation falls short of what the technology could deliver and 34 percent using tools their organisation has not sanctioned; Clio reports fewer than a third of solo and small firms seeing revenue growth from AI; LexisNexis reports 17 percent of UK lawyers saying AI is fully embedded in strategy.
  7. Eight of the thirteen surveys were sponsored by a company that sells legal AI, and in the one pairing that compares like with like (the ABA’s 30 percent and Clio’s 71 percent, two years apart) the vendor’s figure is higher, while ILTA’s 80 percent of firms exceeds LexisNexis UK’s 61 percent of lawyers and the two are not the same quantity. The differences are consistent with question wording and sample construction, and the report treats it as a caveat, not a finding of bias.
  8. No survey in the register asks the question the incident record poses: whether the respondent verifies AI output before relying on it, and how. The nearest is the ABA’s finding that three-quarters of respondents cite AI-generated hallucinations as the reason they have been hesitant.

Why this question

A general counsel deciding on a policy, a regulator writing guidance and a journalist covering the sector all reach for a number: how many lawyers use AI, how many firms govern it. The numbers on offer come from vendors with a product to sell, professional bodies with members to represent, and one UN agency, on samples that range from a hundred firm leaders to two and a half thousand lawyers, with question wording that ranges from “use at least one AI tool” to “AI is fully embedded in strategy”. The search record shows people asking for “legal ai statistics” and “how are law firms using ai” and finding the vendors’ summaries of their own surveys.

The report also has a practical purpose. Backlog item 6.21 of this site’s plan is a survey of its own; this register is the specification of what such a survey would have to ask, and of which existing numbers it would need to be comparable with.

Method and data

A survey qualifies if it was published between 1 September 2024 and 5 September 2026, reports a quantitative result about AI use or governance among lawyers, law firms, in-house departments or judicial operators, and can be read in a document the site has opened (the report, its executive summary, the sponsor’s release or a contemporaneous report by a named outlet). Thirteen qualify. For each, the register records the sponsor and its commercial interest, the publisher, the publication date, the population, the sample size and sampling method as published, the field dates, the geography, the question wording where published, and the headline figures. Where the page read does not state a figure (for example a sample size), the register says so rather than filling it from a secondary source.

The comparability test is simple. Two numbers are comparable if they measure the same quantity (personal use, firm adoption, policy, training), on the same population (private practice, in-house, judiciary), with wording that a respondent would read the same way. By that test, the thirteen surveys yield three comparable pairs and no comparable triple.

Limits. Vendor reports are read from their public pages and press releases; the full instruments are rarely published. Several surveys are annual and this report cites one edition of each. Weighting, response rates and margins of error are not published by any of the thirteen; the report does not compute them. The site’s own earlier page carried four of these surveys; their figures are unchanged. Disclosure: SafeLegalAI is published by Cognesio LLP, whose team also builds LegalAI Space, an AI governance product for law firms; LegalAI Space has published no survey and is not in the register.

The register

#Survey (publisher)PublishedSponsor interestPopulationn and method as publishedField datesHeadline figures
1The Future is Now: AI and the Legal Profession (IBA and CAIDP)30 Sep 2024 (press release; launched at the IBA Annual Conference in September)Professional body and civil-society research centreLaw firms and lawyers, global210 firms responded on policy; lawyer sample not stated on the page readnot stated43% of firms (91 of 210) had any AI policy; 57% said consistency across countries’ rules mattered; 69% unaware how AI regulation would affect their firm
2UNESCO global survey of judicial operators (UNESCO)2024; carried into guidelines launched 4 Dec 2025IntergovernmentalJudges, prosecutors, court staffrespondent count not stated on the pages read; 96 countries202444% had used AI tools in their work; 9% had received any AI-related training or information; 73% wanted mandatory regulations and guidelines
32024 Legal Technology Survey Report (American Bar Association)early Mar 2025Professional bodyPrivate-practice attorneys, US512 attorneys202430% using AI (11% in 2023); 46% of firms with 100+ lawyers, 30% of firms with 10 to 49, 18% of solos; 52% using or considering ChatGPT, 26% CoCounsel, 24% Lexis+ AI; 75% of respondents cite AI-generated hallucinations as a reason for hesitancy
4CLOC 2025 State of the Industry (CLOC, on Harbor’s 2024 law-department survey)2025Industry association; Harbor is a consultancyIn-house legal departments186 departments202430% had an AI tool implemented and in use ; 54% considering implementation within one to two years
5Generative AI survey (LexisNexis UK)1 Sep 2025Vendor (Lexis+ AI)Lawyers, UKnot stated on the releasemid-202561% using generative AI (46% in Jan 2025); 6% no adoption plans (15%); 17% say AI fully embedded in strategy; 51% using legal-sector tools, 70% at medium-sized firms
62025 Technology Survey (ILTA)16 Sep 2025Professional association of legal technologistsLaw firms, mainly US580 firms202580% using or exploring generative AI, from 63% of firms under 50 lawyers to 100% of firms with 700+; 19% no official AI policy (34% in 2024); 45% permit use of vetted or sanctioned tools and applications
72026 AI in Professional Services Report (Thomson Reuters Institute)9 Feb 2026Vendor (CoCounsel, Westlaw)Professionals in law, tax, accounting and others, 27 countriesover 1,500 respondentsnot stated on the page readOrganisation-wide AI use 40% (22% in 2025); 18% know their organisation tracks AI return on investment; 15% of organisations have adopted agentic tools
82026 Legal Industry Report (8am)5 Mar 2026Vendor (practice-management and intake software)Legal professionals, US; 47% partners, 24% paralegals, 12% administrativeover 1,30019 Sep to 18 Oct 202569% use general-purpose AI tools; 46% of firms adopted general-purpose tools and 34% legal-specific; 43% of firms no policy and no plans; 9% written and enforced policy; 54% no training and no plans; 11% mandatory training
92026 Future Ready Lawyer (Wolters Kluwer)10 Mar 2026Vendor (legal research and software)Lawyers in firms and corporate departments, US, China and nine European countries810 quantitative interviews8 to 25 Aug 202592% use at least one AI tool; 62% say AI saves 6% to 20% of weekly time
102026 Legal Trends for Solo and Small Law Firms (Clio)4 May 2026Vendor (practice management)Solo and small firms; geography not stated on the releasenot stated on the releasenot stated71% of solos and 75% of small firms using AI to complete legal work; fewer than a third have increased revenue with AI; 86% of solos made no pricing change
11Future of Professionals Report 2026 (Thomson Reuters Institute)Jun 2026VendorProfessionals in law, tax, audit, accounting, compliance, risk and trade, 62 countries1,816Mar to Apr 202674% use AI several times a week, 44% multiple times a day; 91% say their organisation falls short of what AI could deliver; 34% use unsanctioned tools; 17% say their organisation has no AI strategy; 43% cite untrained people as a barrier
122026 Stand-out Lawyers Survey (Thomson Reuters Institute)6 Aug 2026VendorLaw-firm leaders and managing partners; top lawyers interviewed116 leaders and 2,527 lawyers interviewednot statednearly 80% of stand-out lawyers say their practice has a clear plan for AI; under 50% confident their practice area will succeed; 25% strongly agree the firm has a plan to monetise AI
13The AI-Dependent Lawyer (LexisNexis UK)2 Sep 2026VendorLawyers, UK; large firms and in-houseover 500not stated15% of large-firm lawyers and 13% of in-house lawyers “dependent” on AI to do their job; 34% use AI daily, 32% several times a week

Five adoption numbers that are not the same number

The register’s adoption figures range from 15 to 92 percent. Set side by side with their questions, they sort into five quantities.

Personal use of any AI tool. Wolters Kluwer’s 92 percent (810 lawyers, eleven countries, August 2025) asks whether the respondent uses at least one AI tool, which includes the spell-checker’s successor. 8am’s 69 percent (over 1,300 US legal professionals, autumn 2025) asks about general-purpose tools such as ChatGPT. LexisNexis UK’s 61 percent (September 2025) asks about generative AI in work. These are the high numbers, and they are the ones the sector’s press repeats.

Organisation-wide adoption. Thomson Reuters’ 40 percent (over 1,500 professionals, 27 countries, published February 2026) asks whether AI is used organisation-wide, and its own comparison with 22 percent a year earlier is the cleanest year-on-year figure in the register because the question and sponsor are constant. 8am’s 46 percent (general-purpose tools) and 34 percent (legal-specific tools) are firm-level too.

Use in practice by private-practice lawyers. The ABA’s 30 percent (512 US private-practice attorneys, 2024) is the lowest general-adoption figure and the only one from a bar association. It is also the one with the most transparent firm-size breakdown: 46 percent at firms of 100 or more lawyers, 18 percent among solos. The ABA’s number is a year older than the vendors’ and its population is narrower; both facts pull it down.

Firms using or exploring. ILTA’s 80 percent (580 firms, 2025) counts exploration as adoption. Its size gradient (63 percent under 50 lawyers, 100 percent at 700 or more) is the steepest in the register and the one that matches the incident record, in which the largest firms are the ones that have self-reported.

Dependence and frequency. LexisNexis UK’s 15 percent “dependent” (September 2026) and Thomson Reuters’ 44 percent “multiple times a day” (2026) measure intensity rather than incidence. They are the only figures that would let a reader estimate how much AI-assisted work is reaching courts, and neither is broken down by task.

A reader who wants one number should take the one whose population and question match their own: a US solo practitioner is closest to the ABA’s 18 percent (2024) or Clio’s 71 percent (2026), a gap that measures two years and two sponsors as much as any change in behaviour.

The governance gap, triangulated

Three figures survive the comparability test well enough to triangulate.

Governance measureSource and datePopulationFigure
Firms with any AI policyIBA and CAIDP, Sep 2024210 firms, global43% had a policy
Firms with no official AI policyILTA, Sep 2025580 firms, mainly US19% (34% in 2024); 45% permit use of vetted or sanctioned tools and applications
Firms with no policy and no plans8am, Mar 2026 (fielded autumn 2025)over 1,300 US legal professionals43% no policy and no plans; 9% written and enforced
Judicial operators with any AI trainingUNESCO, 202496 countries; n not stated9% trained; 44% had used AI
Firms providing no training8am, Mar 2026as above54% no training and no plans; 11% mandatory
Professionals citing untrained people as a barrierThomson Reuters, Jun 20261,816, 62 countries43%

The policy figures agree in shape. Fewer than half of firms had any policy in 2024; by 2025 the technology-forward firms ILTA surveys had mostly written one (81 percent), while the broader and smaller population 8am surveys had not (43 percent with no policy and no plan). The enforced-policy figure, 9 percent, is the one a regulator should read: the SRA’s warning notice of August 2026 and the Bar Standards Board’s guidance of May 2026 assume supervision that, on 8am’s number, nine firms in a hundred have written down and applied.

The training figures agree in size and direction across three populations that share nothing else: 9 percent of judicial operators trained (UNESCO), 54 percent of US firms providing none (8am), 43 percent of professionals worldwide naming untrained colleagues as the reason strategy fails (Thomson Reuters). Whichever population is chosen, training trails use by a factor of two to five.

The number that does not exist is the verification rate. No survey in the register asks whether respondents check AI output against the primary source before relying on it, how, or how often. The ABA comes closest by asking about hesitancy (three-quarters of respondents cite hallucination as the reason). The incident tracker’s 150 court decisions are the only measure of what happens when they do not, and it is a measure of detection, not of practice.

Firm size across surveys

Firm size is the one variable that three sponsor-independent surveys and two vendor surveys all report, and it is the one on which they agree.

SurveySmallest firmsMid-sizedLargest firms
ABA 2024 (use in practice)solos 18%10 to 49 lawyers 30%100+ lawyers 46%
ILTA 2025 (using or exploring)under 50 lawyers 63%not broken out on the page read700+ lawyers 100%
LexisNexis UK 2025 (legal-sector tools)not broken outmedium-sized firms 70%not broken out
Clio 2026 (using AI for legal work)solos 71%, small firms 75%mid-market: 51% adjusted pricingenterprise: nearly 60% increased revenue
LexisNexis UK 2026 (dependence)not broken outnot broken outlarge firms 15%

Every survey that breaks out size finds adoption rising with it, and the steepest gradient (ILTA’s 63 to 100 percent) is in the population most likely to have bought an enterprise legal-AI platform. The pattern matches the incident record from the other direction: the largest firms in the sanctions ledger are the ones whose lawyers used a firm-licensed tool (Lacey v State Farm, where the declaration named CoCounsel and Westlaw Precision; Wadsworth v Walmart, where a firm’s internal tool produced eight non-existent cases), while the rows naming ChatGPT are dominated by solos, small firms and litigants in person. Adoption by size and incident by tool are two views of one distribution.

The ABA and Clio solo figures, 18 percent in 2024 and 71 percent in 2026, are the register’s largest apparent change and its least comparable pair: a bar association’s survey of private-practice attorneys asking about use in practice, against a practice-management vendor’s survey of its own market asking about using AI to complete legal work. Some of the gap is two years of adoption. Some of it is the question.

What the questions say

The register’s headline numbers move with wording, and the wording is rarely quoted in coverage. Where the page read gives the question or the category, it is recorded here.

SurveyWhat the respondent was asked, as published
Wolters Kluwer 2026whether they “use at least one AI tool” in their daily work
8am 2026whether they use “general-purpose AI tools” such as ChatGPT, Gemini or Claude for work; separately, whether the firm has adopted general-purpose or legal-specific tools
LexisNexis UK 2025whether they are “using generative AI in their day-to-day work”
Thomson Reuters Feb 2026whether AI is in use “organization-wide”
ABA 2024whether they use AI technology in their practice
ILTA 2025whether the firm is “using/exploring” generative AI tools
Clio 2026whether the firm is “using AI to complete legal work”
Thomson Reuters Jun 2026how often they use AI tools (several times a week; multiple times a day)
LexisNexis UK 2026whether they are “dependent” on AI to do their job; frequency of use
UNESCO 2024whether they had used AI tools in their work; whether they had received any AI-related training or information

Two of the thirteen publish enough of the instrument for the question to be reconstructed (8am, which publishes response categories; UNESCO, which publishes the survey questions). The rest publish a headline and a category label. A reader comparing 92 percent with 30 percent is comparing “at least one AI tool” with the ABA’s category of use of AI technology in practice, and the difference between those two descriptions is most of the difference between the numbers.

Reading the surveys against the record

The surveys measure what respondents say; the incident and regulation trackers measure what courts and regulators wrote down. Set together, three things are visible that neither shows alone.

First, the adoption curve and the incident curve have the same shape and a lag. The vendor surveys put personal use of generative AI at 46 percent in January 2025, 61 percent in September 2025 and 69 percent by the autumn of 2025 (LexisNexis UK and 8am, different populations); the tracker’s incident rows run 11 in 2024, 40 in 2025 and 93 to 3 September 2026. The incident count is partly a function of the tracker’s sweep depth, and the report does not claim a rate from it, but the direction and the timing are consistent with the surveys.

Second, the policy figures explain the sanctions orders’ language. Courts in the ledger sanction lawyers for not verifying, and regulators’ 2026 guidance assumes supervision; 8am’s 9 percent enforced-policy figure and 54 percent no-training figure are the population those orders and that guidance are addressed to.

Third, the surveys’ silence on verification is the gap the courts are filling by decision. Verification is the common floor of the court instruments in the disclosure map that bind filers: 39 of the 46 code it; no survey asks whether anyone does it. The first survey that does will produce the number this register cannot.

Who paid, and what it changes

Eight of the thirteen surveys were sponsored by companies that sell legal AI or legal software: LexisNexis (2), Thomson Reuters (3), Wolters Kluwer, Clio and 8am. Three came from professional bodies (ABA, ILTA, IBA), one from an industry association on a consultancy’s data (CLOC and Harbor), and one from UNESCO.

The sponsor’s interest shows in two places. First, in the population: vendor surveys sample their own customers or panels of professionals reachable through their marketing, and the largest firms, which buy the most software, are over-represented relative to the bar. Second, in the question: “use at least one AI tool” (Wolters Kluwer) is answered yes by more people than the ABA’s category of using AI technology in practice. Where a vendor and a professional body surveyed a comparable population within two years of each other (the ABA and Clio), the vendor’s adoption figure is higher; where the populations differ (ILTA’s firms, LexisNexis UK’s lawyers), the professional body’s figure can be the higher one.

That is not evidence that any figure is wrong. It is evidence that the figures are not interchangeable, and that the only sponsor-neutral figures in the register are the ABA’s 30 percent (2024), ILTA’s 80 percent using or exploring (2025), the IBA’s 43 percent with a policy (2024) and UNESCO’s 44 and 9 percent (2024). Those four are also, by a year or more, the oldest.

The United Kingdom: two vendor surveys and a regulator’s silence

The UK figures in the register both come from LexisNexis: 61 percent of lawyers using generative AI in September 2025 (up from 46 percent that January), and 15 percent of large-firm lawyers “dependent” on it in September 2026, with 34 percent using it daily. Neither release states its sample size on the page read (the 2026 survey is reported as “500-plus”). The Law Society of England and Wales has published guidance (“Generative AI: the essentials”, June 2026) and the SRA a warning notice (17 August 2026), and neither publishes a survey of its regulated population. The Bar Council’s guidance of November 2025 and the BSB’s of May 2026 are likewise unaccompanied by numbers. In the jurisdiction with the deepest incident record outside the United States, the only adoption figures come from a vendor.

In-house and the judiciary

The in-house figure is the oldest and smallest in the register: 30 percent of 186 legal departments had an AI tool implemented and in use in 2024, on Harbor’s survey reported by CLOC in 2025. LexisNexis UK’s 2026 survey reports 13 percent of in-house lawyers “dependent”. Thomson Reuters’ 2026 Future of Professionals survey, which includes in-house respondents, reports that 78 percent of corporate clients consider AI-enabled quality improvements very important or essential while 6 percent say most providers deliver them.

The judiciary has one survey, UNESCO’s, and it is the one whose gap is widest: 44 percent used, 9 percent trained, 73 percent wanting mandatory rules. The guidelines UNESCO launched on 4 December 2025 and the instruments recorded in the bench’s record are the institutional response; no follow-up survey has been published.

What a neutral survey would have to ask

The register specifies its own gap. A survey whose results could be compared with the incident and regulation records would need: a population defined by regulator (solicitors on the roll, members of a bar), not by vendor reach; a sampling frame and response rate published with the results; separate questions for personal use, firm sanction and firm policy; a frequency scale, not a yes or no; a task list that distinguishes drafting, research, summarising and citation-checking; and the verification question, asked directly: before relying on an AI-generated citation, do you open the source, and what proportion of the time. It would report by firm size and jurisdiction, and it would publish the instrument. None of the thirteen does all of these; the ABA and ILTA come closest on population and size breakdown, UNESCO on the training question, and none on verification.

What to watch

The Thomson Reuters organisation-wide series (22 percent in 2025, 40 percent in 2026) is the cleanest year-on-year measure in the register and the one to watch in February 2027. The ABA’s next Legal Technology Survey Report, due in early 2027 on 2026 fieldwork, is the only sponsor-neutral private-practice series. ILTA’s 2026 survey, due in September 2026, will show whether the no-policy figure keeps falling. Any survey that adds the verification question changes the register’s value more than any adoption update.

Three sentences journalists can quote

Thirteen surveys of AI in legal practice published between September 2024 and September 2026 report adoption anywhere from 30 percent (the American Bar Association, private-practice lawyers, 2024) to 92 percent (Wolters Kluwer, lawyers in eleven countries, 2025), because they measure five different quantities for ten different sponsors, eight of them vendors.

The governance figures agree where the adoption figures do not: fewer than half of firms had an AI policy in 2024, 9 percent had one they enforced in late 2025, and 54 percent had provided no training.

No survey in the record asks lawyers whether they verify AI-generated citations before relying on them, the question every court sanction in the SafeLegalAI incident tracker turns on.

Appendix A: the register as data

#SurveySponsor typenAdoption quantityAdoption figurePolicy figureTraining figure
1IBA and CAIDP 2024professional body210 firmsnonenone43% any policynone
2UNESCO 2024intergovernmentalnot statedpersonal use, judiciary44%none9% trained
3ABA 2024professional body512use in practice30%nonenone
4CLOC/Harbor 2024association186 departmentstool implemented30%nonenone
5LexisNexis UK Sep 2025vendornot statedpersonal use, GenAI61%nonenone
6ILTA 2025professional body580 firmsusing or exploring80%19% none; 45% vetted-tools policynone
7Thomson Reuters Feb 2026vendor1,500+organisation-wide40%nonenone
88am 2026vendor1,300+personal use, general-purpose69%43% none; 9% enforced54% none; 11% mandatory
9Wolters Kluwer 2026vendor810any AI tool92%nonenone
10Clio 2026vendornot statedusing AI for legal work, solo and small71% and 75%nonenone
11Thomson Reuters Jun 2026vendor1,816frequency74% weekly, 44% daily17% no strategy43% cite untrained staff
12Thomson Reuters Aug 2026vendor2,643plan and confidence80% clear plannonenone
13LexisNexis UK Sep 2026vendor500+dependence and frequency15% dependent; 34% dailynonenone

Appendix B: sources

Appendix C: changes to this report

  • 5 September 2026: second edition. Expanded from four surveys to a register of thirteen; added the comparability analysis, the sponsor section, the UK, in-house and judiciary sections and Appendix A. The first edition’s four surveys (UNESCO, IBA and CAIDP, ILTA, CLOC) and their figures are retained unchanged as register rows 1, 2, 4 and 6; its note on the UNESCO “160+ countries” misattribution is retained here: that figure belongs to UNESCO’s Global Judges’ Initiative training network, not to the 2024 survey, which covered 96 countries.
  • 16 July 2026: first edition.

Reuse this research

Quote it, cite it, forward it — CC BY 4.0 for the data and figures; the linked official documents are the record. Suggested citation and a link to this exact report: