Notaries: preparing the notarial office for artificial intelligence

Photo: Scott Graham on Unsplash
Artificial intelligence is entering notarial offices through preparatory tasks: research, reviewing documents, first drafts. It does not touch the core of the profession: the authentic instrument (acte authentique) and the responsibility of the person who signs it. Here is what may change, what does not, and what the profession has already set in motion.
What AI can change in the notarial office
The most widespread uses concern the work that comes before the instrument itself. The Banque des Territoires, a long-standing partner of the notarial profession, identifies four [2]:
documentary and case-law research;
analysis and review of contracts and instruments, to spot inconsistencies more quickly;
assisted drafting of instruments and documents based on templates and the data in the file;
processing and automation of time-consuming administrative tasks.
This might mean sorting the documents in an estate file, checking that several documents are consistent with one another or preparing a summary ahead of a client meeting. For the client relationship, the benefit is indirect: the time saved on preparation can be devoted to advice and support [2].
What does not change: the instrument, the signature, the responsibility
An authentic instrument is one that has been received, with the requisite formalities, by a public officer with the authority and capacity to draw it up (Article 1369 of the French Civil Code, Code civil). It is conclusive evidence of what the public officer states that they personally carried out or observed (Article 1371) [4]. This evidential force rests on personal verification. It cannot be delegated to a tool.
The Conseil supérieur du notariat (CSN, the French national council of notaries) is clear on this: AI remains an assistive tool and never replaces the professional, who alone is entitled to take on responsibility for an instrument [2]. Bertrand Savouré, then President of the CSN, reiterated this at the Congrès des notaires de France in autumn 2026: the notary signs the instrument, makes a personal commitment and guarantees its accuracy. In his view, authenticity is regaining 'an entirely new vigour' [3].
The need for verification is all the greater as a result. Generative AI works on a probabilistic basis. It can produce erroneous or approximate content, stated with 'misleading confidence' [2]. Any output must therefore be reread and checked before it goes into a file, and all the more so before it goes into an instrument or a piece of advice.
What the profession is doing
A guide. On 12 December 2025, the Foresight and Digital section of the CSN's Institut d'études juridiques (legal studies institute) published a guide on AI. It sets out three principles: 'absolute respect for professional secrecy', 'control over the processing of personal data' and the notary's 'full and complete responsibility' for the instrument [2].
A framework of trust. On 8 July 2026, the CSN announced an eighteen-month partnership with a French provider of AI models, operated on European cloud infrastructure. It provides for offices to be given access to a conversational agent to help notaries and their staff become familiar with AI, the publication of an AI Charter for the profession, and a joint laboratory to trial notarial use cases. Notaries remain free to choose whether or not to use these tools [1].
Safeguards on data. According to Bertrand Savouré, this partnership comes with commitments on confidentiality, no retraining on the data and no reuse of the data. Secure hosting is being rolled out so that client data remains within the notarial profession [3].
Funded support. The 2027-2030 agreement between the Banque des Territoires and the CSN, presented on 27 October 2026, provides in particular for sovereign hosting of notarial data, stronger cybersecurity and upskilling of notaries in cybersecurity and AI [1].
On the same day, Maître Pierre Jean Meyssan was elected President of the CSN for the 2026-2028 term [1].
Professional secrecy and personal data
A notarial file brings together information on clients' families, assets and sometimes health. Health data is one of the categories of sensitive data, the collection and use of which are prohibited in principle, subject to exceptions [6]. On top of this comes professional secrecy, which the CSN guide places first among its principles [2].
In July 2024, the CNIL (the French data protection authority) published its initial guidance on deploying generative AI within an organisation [5]. It translates directly to a notarial office:
start from a concrete need rather than deploying without a specific purpose;
set rules for use, specifying what is permitted and what is prohibited, for example not providing personal data to the system and not entrusting it with decision-making;
favour local, secure or specialised systems or, failing that, check whether the provider can reuse the data supplied;
train users and raise their awareness of both prohibited uses and the risks of permitted ones;
put in place governance that involves the data protection officer and operational managers from the outset.
The first question to ask, the Banque des Territoires points out, concerns where the data is hosted and what guarantees the provider offers, even before the tool's performance [2].
EU AI Act: training is not optional
Article 4 of Regulation (EU) 2024/1689 on artificial intelligence (the AI Act), on AI literacy, has applied since 2 February 2025 to all providers and deployers, regardless of the system's risk level [8]. Notarial offices that use AI tools in their work are therefore covered.
Regulation (EU) 2026/1744 of 8 July 2026, known as the Digital Omnibus on AI, which entered into force on 27 July 2026 [7], amended this obligation. Providers and deployers must now take measures to promote AI literacy among their staff and the persons using these systems on their behalf, taking into account their knowledge and the context of use. The text no longer requires a specified level to be guaranteed [8]. The obligation remains: national market surveillance authorities have been supervising compliance with it since 2 August 2026 [8].
Two other deadlines deserve attention. Since 2 August 2026, the provider of a system that interacts directly with people must inform them that they are dealing with an AI, which covers, for example, a conversational agent installed on an office's website [8]. The obligations relating to the high-risk systems listed in Annex III have been postponed to 2 December 2027 [8].
Our recommendations
Map the actual uses of AI within the office, including undeclared individual use, and link them to identified needs.
Adopt an internal charter consistent with the profession's AI Charter: authorised tools, data that must never be entered, excluded tasks.
Train notaries and staff, then keep a record of this training (content, participants, dates) with regard to Article 4 of the AI Act.
Introduce a verification rule: no AI output goes into an instrument or a piece of advice without being reread by an identified person.
Review each tool from three angles with your data protection officer: where it is hosted, reuse of data, and compatibility with professional secrecy.
If you would like to review how your office is organised in the face of AI, Tao Sense offers a free diagnostic.
Sources
[1] Press releases — Conseil supérieur du notariat, press releases of 8 July 2026 and 27 October 2026. Accessed 3 November 2026.
[2] IA et notaires : la profession passe à l'acte (AI and notaries: the profession takes action) — Banque des Territoires, 30 September 2026. Accessed 3 November 2026.
[3] Bertrand Savouré : « À l'heure de l'IA, l'authenticité redevient notre atout majeur » ('In the age of AI, authenticity is once again our greatest asset') — Le Monde du Droit, 5 October 2026. Accessed 3 November 2026.
[4] Authentic acts in Europe – France — Notaires d'Europe (Council of the Notariats of the European Union). Accessed 3 November 2026.
[5] Comment déployer une IA générative ? La CNIL apporte de premières précisions (How to deploy generative AI? The CNIL provides initial guidance) — CNIL, 18 July 2024. Accessed 3 November 2026.
[6] Les données sensibles, c'est quoi ? (What is sensitive data?) — CNIL. Accessed 3 November 2026.
[7] Regulation (EU) 2026/1744 of 8 July 2026 (Digital Omnibus on AI) — EUR-Lex (French version), entered into force on 27 July 2026. Accessed 3 November 2026.
[8] AI Act: obligations in force and delays under omnibus 2026/1744 — ActuIA, 6 October 2026. Accessed 3 November 2026.
Comments