Zevra Study · May 2026 (updated)

AI Training for Lawyers: State of play.

Of the 11 mainland French bar schools, only two integrate artificial intelligence in a structured way. Yet the AI Act has imposed AI Literacy since 2 February 2025. On 2 August 2026, in three months, the obligations applicable to the high-risk systems in Annex III take effect.

First published: 14 January 2026 · Last updated: 8 May 2026 · ~ 35 min read

Key figures
77,598 Lawyers registered in France in 2026
4,132 Trainee lawyers across the 11 bar schools
62% Lawyers using generative AI
10,000 Skilia sign-ups in 2 months
2 / 11 Bar schools with a structured AI module
2 Aug AI Act high-risk, full application
Sommaire
Want to train in legal AI?
Discover the Campus Or talk for 30 min

Artificial intelligence has entered the daily life of law firms. According to the CNB's 2024 survey, 62% of French lawyers say they use generative AI in their practice. Skilia, the free training platform launched by the CNB and Lefebvre Dalloz in April 2025, recorded 10,000 sign-ups in two months. The Paris Bar mobilized one million euros to give 14,000 lawyers access to GenIA-L. Lyon, Bordeaux, Marseille and Nancy have launched their own AI plans.

And yet the question of initial training remains open.

Of the 11 mainland bar schools (EDA), only two integrate artificial intelligence in a structured way in their mandatory program: EFB Paris and IXAD Lille. The other nine swing between one-off initiatives and total absence. This means that the majority of the 4,132 trainee lawyers in training in 2026 will leave their school without ever having handled an AI tool in a supervised educational setting.

Lawyers are not only AI users. They will also advise their clients on compliance. A double obligation to get trained.

This study analyzes the AI training offer in French bar schools, bar association momentum, the CNB's work, the European regulatory framework and the landscape of complementary training available. It draws on the official programs published by the bar schools, the minutes of the CNB's general assemblies, bar association communications and data from the profession's Observatory.

Regulatory framework

1. The AI Act: the regulatory framework changes everything

1.1 Application timeline

The European Regulation on artificial intelligence (EU 2024/1689) entered into force on 1 August 2024. Its application is phased:

2 February 2025 In force

Ban on unacceptable AI practices (social scoring, cognitive manipulation, exploitation of vulnerabilities).

2 February 2025 In force

Article 4, AI Literacy: any organization deploying AI must ensure a sufficient level of AI literacy among its staff.

2 August 2025 In force

Obligations for general-purpose AI models (GPAI): transparency, technical documentation, adversarial testing.

2 August 2026 In 3 months

Application to the high-risk systems listed in Annex III: technical documentation, risk management, human oversight, CE marking, registration in the European database. The high-risk systems falling under Article 6(1) — AI embedded in already-regulated products — wait until 2027 (see next line).

2 August 2027 Upcoming

Integration of high-risk systems into already-regulated products (medical devices, machinery, toys).

1.2 High-risk systems in the legal field

Annex III of the regulation classifies as high-risk several categories directly relevant to lawyers:

  • AI systems intended to be used by a judicial authority or on its behalf to research and interpret facts and the law, and to apply the law to a concrete set of facts.
  • AI systems intended to be used to assess a risk relating to natural persons (recidivism, safety risk).
  • AI systems intended to be used for access to and enjoyment of essential public and private services, including legal aid.

The penalties provided for are tiered: up to 35 million euros or 7% of global turnover for prohibited practices, 15 million euros or 3% for non-compliance with the obligations applicable to high-risk systems, 7.5 million euros or 1% for breaches of transparency obligations.

1.3 A double obligation for lawyers

The AI Act creates a double training requirement. The first is direct: Article 4 requires any provider or deployer of AI to take measures ensuring, "to the best of their ability", a sufficient level of AI literacy among their staff. A law firm using ChatGPT, GenIA-L, Doctrine, Jimini or any other AI tool qualifies as a deployer within the meaning of the regulation, and therefore falls within its scope. As the regulation has direct effect, it applies in France without transposition, as of 2 February 2025.

This is an obligation of means: Article 4 sets neither a mandatory curriculum nor a minimum duration. But its flexible wording takes nothing away from its binding character; documented training, tailored to the tools actually used, is the minimum baseline expected by supervisory authorities. Note: Article 4 has no penalty regime of its own in the text; it is the general architecture of Article 99 (up to €7.5M or 1% of turnover for breaches of the other obligations) that applies in the event of a proven failure.

The second requirement is indirect but just as structuring: companies that deploy high-risk AI systems will need specialized legal advice to come into compliance before 2 August 2026. Technical documentation, conformity assessment, risk management, data governance — the AI Act advisory market is an emerging market that trained lawyers will capture. The others will watch the train go by.

The January 2026 study asked the question: are the schools training future lawyers in AI? Europe has since provided the beginnings of a regulatory answer.

Initial training

2. The 11 bar schools: state of play

France has 11 mainland bar schools (EDA) delivering the initial training for the bar admission certificate (CAPA). Add to these 4 to 5 overseas and Corsican structures that deliver only continuing education. The analysis that follows covers the 11 mainland bar schools.

The sources used are the official programs published by each school, the initial-training sections of their websites, institutional communications and articles in the specialist press.

Legend
●Structured AI training ●Partial or targeted initiatives ●No AI module
2.1 · Paris

EFB Paris

Structured AI training

1,860 trainee lawyers. The LAB EFB has been part of the core curriculum since 2018.

The Paris bar professional training school (EFB) is the most advanced bar school on AI integration. It trains roughly 1,860 trainee lawyers each year (the 2025-2026 "Yasmina Reza" cohort), more than 40% of the national total.

The LAB EFB, created in 2018, is built into the core curriculum of the six months of practical foundational teaching. The official 2025-2026 program states that the Lab "familiarizes students with digital technology and more specifically with generative AI through the management of a group project tutored by a specialist instructor". The workshops cover legal design, generative AI, automation and no-code.

The Lab runs semester Pitch Days at Google in partnership with LexisNexis France. The 10th edition was held on 4 December 2025, with the top prize awarded to an AI assistant project for legal departments.

The Talent & Innovation Track is the EFB's 14th selective track. Run by Hercule Legaltech Agency, it offers in-depth specialization in legal tech and innovation.

Since 2024, the EFB has required everyone obtaining the CAPA to have received a minimum of generative-AI literacy.

Verdict · Structured training, built into the core curriculum, with a complete setup (Lab + Pitch Days + selective Track + mandatory literacy).

2.2 · North-West

IXAD Lille

Structured AI training

A dedicated "AI Week" + a usage charter for internships.

The bar practice institute for the Douai Court of Appeal jurisdiction (IXAD) is the second bar school to have integrated AI in a structured way.

The official 2025-2026 program provides for two measures:

  • An AI Week during the program, dedicated to discovering and practicing artificial intelligence tools in a legal context.
  • A generative-AI usage charter framing the use of these tools during firm internships.

This approach combines intensive training (a dedicated week rather than hours scattered around) and ethical framing. It is unique among bar schools outside Paris.

Verdict · Structured training, combining dedicated time (AI Week) and an ethical framework (Charter).

2.3 · Versailles

HEDAC Versailles

Targeted initiatives

8-day "boot camp" at École 42, but reserved for a subgroup.

The HEDAC bar school offers an original setup, but one reserved for a subgroup of students.

The 2026-2027 program includes an 8-day "boot camp" immersion at École 42 for trainee lawyers enrolled in the Intellectual Property & New Technologies specialization, as part of the "Law, Entrepreneurship and Digital" master's co-accredited with Paris-Saclay and HEC. An intensive format, but one that concerns only a fraction of the cohort.

The HEDAC also offers a "Digital transformation of law & legaltech" university diploma from Paris II Panthéon-Assas, accessible through a partnership, and organizes conferences as part of the "Wednesdays at HEDAC" series.

Verdict · Significant but targeted initiatives (IP/Tech subgroup). No generalized AI module.

2.4 · Greater West

EDAGO Rennes

Awareness in the program

Digital-tools awareness + an annual AI symposium.

The Greater West bar school (EDAGO), based in Bruz near Rennes, covers the Rennes, Angers, Caen and Nantes jurisdictions.

The official 2025-2026 program contains two notable elements: the practical workshops include "awareness of the use of various digital tools", and the program provides for a symposium on artificial intelligence at the end of the school year, co-organized with students and open to continuing education.

The specialization tracks focus on classic legal subjects; there is no dedicated "digital" or "innovation" track.

Verdict · Digital-tools awareness and an annual AI symposium written into the official program. More advanced than average but short of a structured module.

2.5 · Bordeaux

EDA Aliénor Bordeaux

Continuing education only

No AI module in initial training, but momentum in continuing education.

EDA Aliénor (Bordeaux) offers no identifiable AI module in its initial training program.

It is, however, active in continuing education. An AI cycle led with Pierre-Xavier Chomiac de Sas was run over several sessions (March 2024, June 2024, March 2025), centered on ChatGPT and lawyers' practice. The back-to-school university on 29-30 August 2025 in Lège-Cap-Ferret had as its theme "Lawyers and artificial intelligence".

The school is part of the momentum of the Bordeaux Bar, positioned as an "AI Ambassador" in July 2025.

Verdict · No AI module in initial training. Significant momentum in continuing education.

2.6 · Toulouse

EFA Toulouse

Events

6th CNB Forum 2025 on AI, but nothing carried over into the curriculum.

The Toulouse bar school organized the 6th CNB Training Forum on 15-16 May 2025, whose central theme was innovation and artificial intelligence. The panel brought together Yann Ferguson (INRIA), with workshops from Lefebvre Dalloz, Lexbase, LexisNexis and Septeo.

Despite this success, the official initial-training program contains no dedicated AI module. The Forum is a triennial event (next in 2028).

Verdict · A major institutional event (CNB Forum 2025) but nothing carried over into the initial-training program.

2.7 · Rhône-Alpes

EDARA Lyon

No AI module

AI momentum plays out at the Lyon Bar, not at the school.

The Rhône-Alpes bar school (EDARA), based in Lyon, covers the entire Lyon Court of Appeal jurisdiction. The official program mentions internships, work-study and the fundamentals of professional ethics. No AI module is identified in the initial curriculum.

The jurisdiction's AI momentum plays out at the level of the Lyon Bar, which launched an "AI Action Plan 2026" on 3 March 2026 with monthly AI Cafés and partnerships with Doctrine, Lexbase and Side Quest.

Verdict · No AI module in initial training. Momentum driven by the Lyon Bar.

2.8 · South-East

EDASE Marseille

No AI module

Continuing-education partnerships via the Marseille Bar.

The South-East bar school, based in Marseille, covers the Aix-en-Provence and Nîmes jurisdictions. The initial-training program contains no identifiable AI module.

In continuing education, the Marseille Bar has struck a partnership with Lexbase (free, unlimited access) and schedules in-person training at the Maison de l'Avocat.

Verdict · No AI module in initial training. Continuing-education partnerships via the Bar.

2.9 · Greater East

ERAGE Strasbourg

AI as an elective subject

AI offered as an elective subject in the formal curriculum.

The Greater East regional bar school (ERAGE), based in Strasbourg, covers Colmar, Metz, Nancy, Dijon, Besançon and Reims. With around 200 trainee lawyers per cohort and 3,700 registered lawyers.

The official program offers artificial intelligence as an elective subject for students to choose, alongside mediation, environmental law and local law. This classification places AI within the formal curriculum; it is an identified elective module in the teaching offer. ERAGE also mentions a "trainee lawyer digital charter" that students must sign upon enrolment.

The Nancy Bar, which is part of the jurisdiction, has collectively subscribed to GenIA-L (Lefebvre Dalloz) for all its lawyers.

Verdict · AI built in as an elective subject in the curriculum, which places it above average. Formalized digital charter.

2.10 · Central-South

EDACS Montpellier-Clermont

No AI module

Rich local legaltech ecosystem, but the school does not tap into it.

The Central South bar school (EDACS), based in Montpellier with a branch in Clermont-Ferrand, offers no identifiable AI module in its initial curriculum.

Montpellier hosts a dense legaltech ecosystem (Septeo, LegalPlace, a Legaltech & Law Innovations university diploma co-certified with Stanford CodeX). The Montpellier Bar has nonetheless launched a 34-hour AI & Data Certification, one of the most ambitious continuing-education AI programs offered by a local bar.

Verdict · No AI module in initial training despite a rich local legaltech ecosystem. Strong momentum driven by the Montpellier Bar.

2.11 · Central-West

ECOA Poitiers

No AI module

No AI module in the published program.

The Central-West bar school (ECOA), based in Poitiers, covers the Poitiers, Limoges, Bourges, La Rochelle, Saintes, Niort and Angoulême jurisdictions. No AI module is identified in the published initial-training program.

Verdict · No AI module in initial training.

Summary

Map of the 11 bar schools

School Jurisdiction Status
EFB Paris Paris Structured AI training
IXAD Lille North-West Structured AI training
HEDAC Versailles Versailles Targeted initiatives
EDAGO Rennes Greater West Awareness in the program
EDA Aliénor Bordeaux Bordeaux Continuing education only
EFA Toulouse Toulouse Events
EDARA Lyon Rhône-Alpes No AI module
EDASE Marseille South-East No AI module
ERAGE Strasbourg Greater East AI as an elective subject
EDACS Montpellier-Clermont Central-South No AI module
ECOA Poitiers Central-West No AI module
Professional initiatives

3. Bar association momentum: the invisible movement

Analyzing the bar schools' programs does not capture the real momentum. The most significant movement plays out at the level of the bar associations and continuing education, not initial training.

3.1 Paris Bar: a one-million-euro plan

The Paris Bar published its first white paper on artificial intelligence in October 2025, presented as the first ethical framework of this scale developed by a French bar. The document includes a best-practices guide, an FAQ and a checklist, and is designed as a living document.

In parallel, the Bar mobilized one million euros to fund free access to GenIA-L (Lefebvre Dalloz) for around 14,000 Parisian lawyers practicing solo or in pairs, until 31 December 2025. The training included two mandatory 30-minute sessions (professional ethics and professional secrecy). A preferential rate was negotiated for 2026 (€80/month for GenIA-L for Search, €220/month for GenIA-L Assistant).

In total, seven strategic partnerships have been struck since October 2024: Lefebvre Dalloz/GenIA-L, Doctrine, Jimini, Pappers Justice, Ordalie, Lexis+ AI/JP Intelligence in the library. These partnerships cover more than 9,000 lawyers.

In January 2026, the Bar launched Themisia, a firm-management software co-built with Xelya and made freely available to the 35,000 Parisian lawyers, tied to the electronic-invoicing obligation of September 2026.

3.2 Bordeaux: AI Ambassador

The Bordeaux Bar positioned itself as an "AI Ambassador" in July 2025, as part of the national "Dare AI" plan. Four partnerships were signed simultaneously: Haiku, Doctrine, LexisNexis and Lexbase. The Bordeaux lawyers' back-to-school university (August 2025) was devoted to AI.

3.3 Lyon: AI Plan 2026

The Lyon Bar launched an "AI Action Plan 2026" on 3 March 2026 during an "Artificial intelligence: from literacy to integration" day at the Campus Région du numérique. The plan includes AI Cafés (16 sessions scheduled in 2026, on the 1st and 3rd Friday of each month) and partnerships with Doctrine, Lexbase and Side Quest.

3.4 Other bar associations

  • Nancy · collective subscription to GenIA-L for all its lawyers.
  • Marseille · Lexbase partnership with free, unlimited access, in-person training at the Maison de l'Avocat.
  • Strasbourg · Lexbase "Using legal AI" workshops (March 2026).
  • Saint-Étienne · Lexbase partnership and a joint study with Cercrid and the University of the Antilles.
  • Grenoble · Lexbase AI partnership.
  • Montpellier · a 34-hour AI & Data Certification (10h e-learning + 24h hybrid). Co-directed by Malo Depincé (University of Montpellier) and Lyoma Kogiso (lawyer). One of the most complete programs offered by a local bar.

3.5 What this map reveals

The bar associations are moving. The bar schools, for the most part, are not. The gap is all the more striking because continuing education follows an individual logic (the lawyer chooses whether or not to get trained) whereas initial training is mandatory and structured by the schools.

The question is not whether lawyers can train themselves in AI — the offer exists, it is abundant, it is sometimes free. The question is whether the training of future lawyers prepares them for the world in which they will practice.

Institutional

4. The CNB: from observation to action

4.1 Timeline of the CNB's AI work

The National Bar Council has significantly accelerated its work on AI since mid-2025.

April 2025

Launch of Skilia Avocats, a free 4-hour training valid for continuing-education credit. 10,000 sign-ups in two months.

23 May 2025

Great Consultation: 25% of lawyers take part. Digital and AI topics are ranked the 2nd priority project of the profession.

13 June 2025

AI progress-review general assembly. Publication of a guide to choosing legal AI tools with an analysis grid (sovereignty, security, reliability, intellectual property). The CNB is heard by the Senate and is in dialogue with the Chancellery, the Council of State and the Court of Cassation.

10 April 2026

General assembly: an AI clause is added to the standard fee-agreement templates. The first concrete normative translation. Lawyers are now invited to inform their clients of their use of AI.

4.2 Skilia: the right answer to the wrong question?

Skilia is a foundation. Four hours of free e-learning, valid for continuing-education credit, accessible to all: that is a genuine gain. But Skilia covers the theoretical fundamentals, not operational practice. Knowing what AI is is not the same as knowing how to use it in a defense brief, a due diligence or a case-law search.

The stakes lie beyond awareness. They lie in practical mastery of the tools, the critical reflex when faced with model outputs, the integration of AI into the firm's processes without compromising confidentiality or professional ethics. Skilia lays the first brick. Others are needed.

4.3 CAPA apprenticeship contract: voted but not yet effective

The CNB unanimously voted on 11 April 2025 to open the apprenticeship contract to trainee lawyers. Entry into force is formally set for 1 January 2026, but it is optional in the volunteering CRFPAs and conditional on the CAPA being registered in the RNCP. The CNB itself indicates that "this rollout will probably not be effective for the benefit of the 2026-2027 cohort". The CRFPAs' unified internal rules target the 2027-2028 cohort.

No AI component is built into the skills framework or into the training of apprenticeship supervisors.

Professional practice

5. AI in legal practice: risks, technologies and use cases

The preceding sections paint an institutional picture. This one gets concrete. A lawyer interested in AI asks three questions: what are the real risks, what technologies exist, and what can you do with them in a firm?

5.1 Generative and non-generative AI: a distinction worth knowing

The most widespread confusion among lawyers is to reduce artificial intelligence to ChatGPT. Legal AI in fact covers two very different families of technology.

Non-generative AI has been present in legal tools for years. Case-law search engines (Doctrine, Lexbase, LexisNexis), contract-analysis tools (Hyperlex, Legartis, Jus Mundi), legal-analytics and judgment-prediction platforms (Predictice, Case Law Analytics). These tools "create" nothing. They analyze, sort, classify and return existing information. Their main risk is statistical bias.

Generative AI — ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Mistral, Copilot (Microsoft), GenIA-L (Lefebvre Dalloz) — works differently. These large language models produce original text from statistical patterns learned on massive corpora.

A non-generative case-law search engine returns decisions that exist: the risk is omission, not invention. A generative model can return a decision that never existed, with a credible docket number and a perfectly plausible holding.

5.2 Hallucinations and bias: the two major risks

Hallucinations are the most publicized and most misunderstood risk. A generative language model does not "lie". It produces the statistically most probable continuation of text, regardless of its truthfulness. In law, this translates into invented case-law references, non-existent code articles, reasoning that is legally coherent but factually false.

The documented cases are numerous. The Mata v. Avianca case (United States, 2023) revealed that briefs drafted with ChatGPT cited six non-existent decisions. In the United Kingdom, the Harber v. AI Solutions case (2024) led to disciplinary sanctions. In France, no public sanction has yet been handed down, but the CNB's Rules and Practices committee has been working on it explicitly since June 2025.

The hallucination rate varies considerably across models. Published estimates range between 3 and 15% for the most recent models under optimal conditions, and far more on precise legal questions. The remedy is not to avoid using AI. It is to systematically verify every reference, every citation, every piece of reasoning.

Bias is a more insidious risk. AI models are trained on corpora that reflect the biases existing in legal literature and case law. A model fed historical case law will reproduce the biases of gender, origin or social class present in past decisions.

For a lawyer, mastering these risks comes down to three reflexes: understand the type of model used, never treat an AI output as a reliable source without verification, and document your control process.

5.3 Assessing risks and opportunities: four dimensions

  • Confidentiality · Consumer-grade solutions (ChatGPT Free, Gemini Free) offer no sufficient guarantee for professional use involving data covered by professional secrecy. Professional solutions (ChatGPT Enterprise, Claude for Work, GenIA-L, Mistral API) offer contractual commitments.
  • Reliability · Exploratory research is a low-risk, high-gain use case. Drafting documents that engage the lawyer's liability is high-gain but high-risk if human oversight is insufficient.
  • Professional ethics · The AI clause in the standard fee agreements (CNB general assembly, April 2026) formalizes the requirement of transparency toward the client.
  • Cost/benefit · Is the investment in training, subscriptions and oversight time offset by productivity gains?

5.4 Concrete use cases

AI does not transform legal practice in the abstract. It transforms it task by task.

Low risk, immediate gain
  • Exploratory case-law research · 30 to 60 min saved per search.
  • Summarizing large documents · 1 to 3 hours saved per document.
  • Preparing client meetings · 15 to 30 min per meeting.
  • Legal translation · quality far above classic tools.
Moderate risk, structural gain
  • Drafting first drafts · gain on the blank-page time.
  • Contract analysis and contract management · specialized tools (non-generative AI) are more reliable.
  • Automated legal monitoring · regularity without a daily time burden.
Advanced use cases
  • Workflow automation · generation of standard documents, extraction of information from exhibits. Prompt-engineering skills become decisive.
  • Access to databases via MCP (Model Context Protocol) · direct querying of Légifrance, JURI, collective agreements, JORF, with sourced and verifiable answers.
  • Building custom tools · vibe coding now lets a non-developer lawyer create functional applications.
Market landscape

6. Landscape of AI training for lawyers

Beyond the bar schools and Skilia, a market for legal AI training has taken shape.

6.1 Comparison of available training programs

Training Duration Price Format Qualiopi
Bootcamp AI & Legal Tech Zevra Zevra Campus · Practicing lawyers 3 days €3,400 excl. VAT In person (Bordeaux) ✓
Legal AI training Side Quest · Lawyers (all levels) 3h to 1 day €200–350 In person / video ✓
AI for Lawyers Lefebvre Dalloz · Lawyers 1 day €1,039 excl. VAT In person (Paris) ✓
EDHEC ALLL1 Certificate EDHEC / AFJE · In-house counsel + lawyers 12h ~€1,500 100% video ✓
Generative AI Bootcamp for Lawyers Le Juriste de Demain · Lawyers 3 days N/A 100% video ✓
Skilia Avocats CNB / Lefebvre Dalloz · All lawyers 4h Free E-learning –
AI for law firms ENADEP · Lawyers / assistants 4h30 N/A Video ✓
Legal AI training Gaius · Lawyers / in-house counsel Variable Variable Video ✓
AI in the legal function Paris 1 Sorbonne · In-house counsel + lawyers 7h N/A In person (Paris) –
AI for legal & compliance Telecom Paris · Legal + compliance 2 days N/A In person (Paris) ✓

FIF-PL eligibility: self-employed lawyers can tap their FIF-PL fund to finance their continuing education. Qualiopi-certified training is eligible.

6.2 What this landscape reveals

The offer exists. It covers every format (e-learning, video, in person, bootcamp), every budget (free to €3,400) and every level. Side Quest reports 750 lawyers trained with a recommendation rate of 97.2%.

The paradox is striking: a market for legal AI training exists, it is dynamic, it answers a real demand, but it remains external to the initial curriculum of 9 bar schools out of 11.

6.3 The real economics: the cost of inaction

A lawyer who wants to train seriously in AI can stack short courses: a half-day of awareness here (€200-350), a hands-on day there (€1,039), a video certificate elsewhere (~€1,500). By combining three or four courses, they will have invested €2,500 to €3,500 and committed several days scattered over several months. No educational continuity.

The alternative is the immersive format. Zevra Campus's Bootcamp Vibelawyer runs on this model: three consecutive days in person in Bordeaux, around fifteen participants, a progression that starts from the fundamentals and reaches operational mastery. The price, €3,400 excl. VAT, includes one year of access to Zevra Campus.

A lawyer billing €200 excl. VAT per hour who saves 30 minutes a day recovers ~€2,500 of billable work per month. Payback in six weeks.

What Bootcamp participants report systematically is the value of being in person. Not for the content — you can learn the same concepts over video. For everything else. The pair work, the informal exchanges between practitioners, the real-time resolution of technical roadblocks. When a criminal lawyer and a tax lawyer work side by side on the same tools, each leaves with use cases they would never have thought of alone.

FIF-PL eligible · Qualiopi certified.

6.4 The "after" problem

Three days of intensive training do not solve everything. AI evolves. The tools change. What works in May 2026 will not work the same way in December. The bootcamp-trained lawyer who does not maintain their skills becomes obsolete in six months.

The challenge is not only to train lawyers in AI. It is to keep them trained.

International

7. European benchmark

7.1 Germany

As early as 2024, Bavaria integrated a "Legal Tech & AI" module into the Rechtsreferendariat (post-studies legal traineeship). The initiative remains isolated. The other Länder have not followed. The universities (Munich, Hamburg, Bucerius Law School) offer elective courses but no mandatory module in the training of Rechtsanwälte.

7.2 United Kingdom

King's College London launched an "AI Literacy for Lawyers" program built into the LLM curriculum in 2024. The SRA published guidance on the use of AI by solicitors in 2025. Yet no AI-training obligation exists within the SQE framework.

7.3 CCBE

The Council of Bars and Law Societies of Europe, at the CNB general assembly of 10 April 2026, stressed the impact of AI on the profession and the importance of strengthening training programs at the European level. Recommendations on AI and professional ethics are in preparation.

7.4 What it says

Nobody really has a head start. France is in the European average. But the initial-training deficit is not a French exception. It is a common structural problem: everywhere in Europe, lawyer-training curricula evolve slowly. Technology, for its part, does not wait for them.

Conclusion

8. Three months to change paradigm

On 2 August 2026, the AI Act takes full effect for high-risk systems. Article 4 on AI Literacy is already in force. The CNB has added an AI clause to the standard fee agreements. The Great Consultation ranked AI as the profession's 2nd priority.

The movement is real.

But it still rests largely on individual goodwill. The lawyer who wants to get trained can do so — the offer exists, it is abundant, it is sometimes free. The lawyer who does not want to can obtain their CAPA without ever having opened an AI tool in an educational setting. That is the case in 9 bar schools out of 11.

The question is no longer whether AI will transform the profession. It is already transforming it. The question is whether the bar schools will prepare the next cohorts for this reality, or whether they will keep training lawyers for a world that will no longer exist when they are sworn in.

Cette étude dresse un état des lieux. Si votre besoin est plus immédiat, notre catalogue de formation IA Avocat présente neuf formations, du parcours gratuit au programme sur mesure. Pour arbitrer entre les offres du marché avant de choisir, nous avons publié à part un guide des critères de choix : formats, prix, financement et validation des heures.

Zevra solution

9. Zevra Campus: continuing education as infrastructure

If the bar schools do not structurally train for AI, and if one-off courses solve the need of the moment without guaranteeing lasting upskilling, then the question that arises is one of infrastructure. Zevra Campus was designed to answer it.

The platform rests on three pillars: a structured e-learning program ("AI Fundamentals", 12 video capsules), regular webinars led by practitioners, and community access for peer-to-peer exchange.

FREE

Rookies account

€0 Access to the entire e-learning track + a selection of webinars. No credit card.
EARLY ADOPTER

Campus account

€1,250 excl. VAT/yr €2,900 · or 3×€500. Advanced webinars, replays, workshops, community.

The Bootcamp Vibelawyer (section 6.3) includes one year of Campus access. Three days of immersion to spark the momentum, then twelve months to keep it going.

For structured firms, the Campus works on a multi-user license. The Campus does not claim to replace the bar schools. It fills a gap they leave open.

Discover the Campus See the Bootcamp
Frequently asked questions

FAQ

Is there a legal obligation for lawyers to train in AI?
Yes, since 2 February 2025. Article 4 of the AI Act requires any organization deploying AI to ensure a sufficient level of AI literacy among its staff. A law firm using AI tools falls within the scope of this obligation. On the other hand, no specific obligation exists within the CAPA curriculum.
What is the best AI training for a lawyer?
It depends on where you start and what you're looking for. Skilia (CNB, free, 4 hours) lays the groundwork. One-day courses (Lefebvre Dalloz, Side Quest) allow a first hands-on practice. The 3-day bootcamps (Zevra Campus, Le Juriste de Demain) aim for operational mastery.
How much does AI training for a lawyer cost?
From €0 (Skilia CNB) to €3,400 excl. VAT (intensive bootcamp). Most Qualiopi-certified courses are eligible for FIF-PL funding.
Is the CNB's Skilia training enough?
Skilia covers the fundamentals in 4 hours of e-learning. It's a useful foundation. But knowing what AI is is not the same as knowing how to write an effective prompt for a case-law search, or how to automate a firm workflow. Supervised practice is another matter.
Which bar schools train in AI?
EFB Paris (Lab EFB, Talent & Innovation Track) and IXAD Lille (AI Week, usage charter) are the only two mainland bar schools with a structured, mandatory AI setup. ERAGE Strasbourg offers AI as an elective subject. HEDAC Versailles and EDAGO Rennes offer significant initiatives but ones reserved for subgroups or one-off formats.
What is the AI Literacy imposed by the AI Act?
Article 4 of EU Regulation 2024/1689 requires AI providers and deployers to ensure their staff have a "sufficient level of AI literacy". For a law firm, this means training associates, partners and trainees in the use of these tools while respecting professional ethics and data confidentiality.
What's the difference between the Bootcamp Vibelawyer and the Zevra Campus?

The Bootcamp is 3 days in person (€3,400 excl. VAT, FIF-PL eligible). You gain operational mastery. You leave with one year of Campus access included.

The Campus is a continuing-education platform over 12 months. Free Rookies account (no credit card), Campus account at €2,900 excl. VAT/yr (or €1,250 excl. VAT Early Adopter). E-learning, webinars, community. The Bootcamp sparks the momentum, the Campus keeps it going.

Can you start for free?
Yes. Two options. The Zevra Campus Rookies account is free, with no credit card: access to the "AI Fundamentals" (9 video capsules) track and a selection of webinars. Skilia (CNB) is also free (4 hours of awareness). The two are complementary.
How do you write effective prompts?

An effective legal prompt rests on four elements: the role (telling the model to act as a specialized lawyer), the context (describing the factual situation precisely), the instruction (stating what you expect), and the output format (asking for a structure: outline, table, list).

The more precise the prompt, the less the model improvises. The cardinal rule: a prompt is not a question typed into Google. It is a brief given to a very fast associate who has no judgment.

How do you measure the ROI of AI in a firm?

Three indicators. Time saved per task — most trained lawyers report 30 minutes to 2 hours saved per day. Billing recovered — 45 minutes/day at €200/h = ~€3,300/month. The qualitative side — depth of analysis, thoroughness, speed.

The total annual cost for an individual lawyer is between €2,000 and €5,000. The break-even point is generally reached in 4 to 8 weeks.

And some gains are immediate. A lawyer trained in vibe coding can, in a week, build their own website and run it alone in a few minutes a week. Real gain compared to hiring an agency: more than €5,000 over three years, not counting the autonomy gained over updates, SEO and publishing content.

How do you govern the use of AI in your firm?

Four pillars:

  • Usage policy · Which tools, for which tasks, with which data.
  • Control process · Any AI output in a client deliverable must be checked by a qualified professional.
  • Client transparency · The AI clause in fee agreements (CNB general assembly, April 2026) is its normative translation.
  • Continuing education · For organizations of more than five people, appointing an internal AI lead is a high-leverage investment.
Sources

Sources and methodology

This study analyzes the official programs of the 11 mainland bar schools (annual PDF documents published on the schools' websites), the minutes of the CNB's 2025 and 2026 general assemblies, bar association communications, training providers' websites and data from the national Observatory of the legal profession. The information is verified as of 8 May 2026.

Regulatory texts

AI Act (EU Regulation 2024/1689), Full text in the OJEU

CNB, general assemblies and AI work

Key figures 2026 of the legal profession, cnb.avocat.fr
Directory of bar schools (EDA), cnb.avocat.fr
General assembly of 13 June 2025, AI progress review, cnb.avocat.fr
Launch of Skilia Avocats (April 2025), cnb.avocat.fr
Training Forum 2025, Toulouse, cnb.avocat.fr
Choosing your bar school for 2026, cnb.avocat.fr
The Observatory of the legal profession, cnb.avocat.fr

Bar school programs

EFB Paris, 6 months of foundational teaching, wpvit.efb.fr
EFB Paris, Talent & Innovation Track, wpvit.efb.fr
EFB Paris, Actu-Juridique interview, actu-juridique.fr
IXAD Lille, ABC of initial training, ixad.fr
HEDAC Versailles, 2026-2027 program (PDF), hedac.fr
EDAGO Rennes, 2025-2026 program (PDF), edago.fr
EDARA Lyon, edara.fr
ERAGE Strasbourg, Initial training, erage.eu
EDACS Montpellier-Clermont, edacentresud.com
EDASE Marseille, Initial training, edase.fr
ECOA Poitiers, Initial training, fi.avocats-ecoa.fr

Bar associations, AI initiatives

Paris Bar, white paper on AI (Oct. 2025), avocatparis.org
Paris Bar, AI: the bar in action, avocatparis.org
Bordeaux Bar, AI Ambassador, barreau-bordeaux.com
Lyon Bar, AI: from fear to strategy (March 2026), mesinfos.fr
Marseille Bar, EDASE training, barreau-marseille.avocat.fr
Strasbourg Bar, Training, avocats-strasbourg.com
Montpellier Bar, AI & Data Certification (34h), barreau-montpellier.com

Demographic data

Statistics on the legal profession 2023, justice.gouv.fr (PDF)
Key figures of the legal profession (Yves Avril), yvesavril.com

CAPA apprenticeship contract

The apprenticeship contract opens to lawyers in 2026, alternance-professionnelle.fr
Trainee lawyer and apprentice: it's voted, labase-lextenso.fr
Registration of the CAPA in the RNCP, CNB, cnb.avocat.fr

Case law on AI and lawyers

Mata v. Avianca, Case No. 22-cv-1461 (PKC), United States District Court, Southern District of New York, 2023.

Zevra is a legaltech development studio based in Bordeaux. Zevra Campus offers AI training for lawyers, including the Bootcamp Vibelawyer. Side Quest, the training arm of the Zevra group, has trained more than 750 lawyers. This study is published in the interest of transparency about the state of AI training in the profession.

The authors

The team behind the study

This study is produced by the Zevra team, who train lawyers in AI and build production legal tools every day.

Pierre Colliot

Pierre Colliot

Président · Chief AI Officer

11 ans d'innovation dans le secteur juridique. Fondateur de La Fabrique Juridique, créateur de Supernovia.io, ex-Directeur Marketing Jarvis Legal. A accompagné des dizaines de cabinets et legaltechs dans leur croissance.

Alexis Deborde

Alexis Deborde

Co-fondateur

Juriste devenu entrepreneur, pionnier de la legaltech française. Fondateur de Leganov, co-fondateur d'Hercule, créateur de Smartpreuve. Classé parmi les « 40 qui font le futur du droit » par l'Institut Choiseul.

Joachim Darmalingon

Joachim Darmalingon

CTO

7 ans en développement d'applications SaaS. Ex-développeur Legal Pilot, Lead Dev Supernovia.io, créateur de Lexiwork et KaribNav. Ingénieur MIAGE, spécialiste des architectures scalables.

Benjamin Barbé

Benjamin Barbé

Infra & Cybersécurité

Tech Lead et Solution Architect, fondateur de Techspike, créateur de la Zevra Factory. Spécialiste des migrations complexes, architectures distribuées et cybersécurité juridique (RGPD/NIS2, ANSSI).

Pierre-Louis Roquet

Pierre-Louis Roquet

Avocat · Animateur Campus

Avocat au barreau de Lyon, co-fondateur de SmartLawyer. Anime la communauté Zevra Campus et intervient sur les formations Avocats. Un Avocat qui forme des Avocats.

© 2026 Zevra. All rights reserved. Legal notice Privacy Terms DPA