Legal technology and the practice of law
AI in Zurich law firms
The legal profession in Zurich, which serves the financial centre as advisers, litigators and compliance specialists, has adopted AI tools with growing speed. The large commercial firms use platforms for document review in due diligence, contract analysis, translation and the preparation of first drafts. Smaller firms rely on general-purpose assistants and on specialised Swiss legal databases that have added AI search functions. The Zurich start-up DeepJudge, described in section III, is an example of a local company that has built its business on the specific need of law firms to search their own confidential knowledge without exposing it.
The Swiss Bar Association (SAV/FSA) adopted guidance on the use of artificial intelligence on 14 June 2024. It names as typical applications translation software, data analysis in internal investigations, client due diligence and the use of generative AI to summarise and edit texts.1 The guidance requires lawyers to clarify before using any software, including AI applications, what happens to the data they enter. It describes three compliant approaches: solutions operated on the firm’s own premises, arrangements that meet the requirements for outsourcing, and informed client consent. Confidential information and business secrets must not be entered into AI systems that do not meet these conditions. The guidance stresses that results must be reviewed independently and critically, because they may contain hallucinations, gaps or bias, and it states that the lawyer remains responsible for the quality of the work.2
Professional secrecy under Article 321 SCC
The central legal constraint is professional secrecy. Article 321 of the Swiss Criminal Code makes it an offence for lawyers and their auxiliaries to disclose secrets entrusted to them in their professional capacity, while Article 13 of the Federal Act on the Free Movement of Lawyers imposes the same duty as a professional rule.3 Whether the use of a cloud service or an AI tool constitutes a disclosure depends on the technical arrangement. An expert opinion commissioned by the Swiss Bar Association and prepared at the University of Zurich in 2019 concluded that no disclosure occurs where the lawyer encrypts the data before transmission so that the provider cannot read them. Where the provider itself encrypts the data after receipt, the provider may be treated as an auxiliary within the meaning of Article 321 SCC, provided that it is carefully selected, contractually bound to confidentiality and permitted to use the data only to perform the contract. On this basis the Bar Association regards cloud computing as permissible in principle, subject to a risk assessment for each mandate, a contractual secrecy obligation, client consent where appropriate and particular caution with foreign providers because of the risk of access by foreign authorities.4
Applied to AI, this analysis produces a graded picture. Using a public consumer chatbot with client names and facts is incompatible with professional secrecy, because the provider may use the input for its own purposes and is not bound as an auxiliary. Enterprise versions of the same models, contractually excluded from training on client data and bound to confidentiality, can meet the auxiliary test if the other conditions are fulfilled. Models operated within the firm’s own infrastructure raise no disclosure issue at all. For international clients, the practical consequence is that Zurich law firms increasingly explain in their engagement letters which AI tools they use and on what terms. They also obtain consent where the arrangement goes beyond what the auxiliary doctrine clearly covers.
Liability, fees and the future of legal work
The professional rules also shape how the benefits of AI are shared. A lawyer who uses AI to produce a first draft in minutes must still review it with the same care as a draft prepared by a junior associate, and the lawyer bears full responsibility for errors. Fee arrangements based on hourly rates come under pressure when tasks that once took hours take minutes, which favours fixed fees and value-based pricing. The SAV guidance does not address fees in detail, but it reminds lawyers that clients may expect work to be performed personally and that a duty to inform may arise if AI replaces personal work in a way the client would not anticipate.
For the financial centre, the legal profession’s use of AI matters in a specific way. Banks and asset managers in Zurich rely on external lawyers for regulatory advice, cross-border questions and disputes. If those lawyers can review larger volumes of documents more quickly and at lower cost, the financial centre as a whole becomes more efficient. The same applies to compliance functions within the institutions, many of whose members are trained lawyers. The limiting factor is the capacity of experienced professionals to verify the output of the technology. The scarcest resource is people who understand both the law and the machine, which leads back to the theme of this chapter.
Notes
- Swiss Bar Association, “Wegleitung für die Anwaltschaft für den Umgang mit KI”, https://www.sav-fsa.ch/news/-/asset_publisher/97aPCN6lT9LD/content/wegleitung-k%C3%BCnstliche-intelligenz-ki- (accessed September 2026). ↩
- Swiss Bar Association, “SAV-Wegleitung für den Umgang mit künstlicher Intelligenz”, version published February 2025, https://digital.sav-fsa.ch/documents/1060627/0/SAV-Wegleitung_f%C3%BCr_den_Umgang_mit_k%C3%BCnstlicher_Intelligenz_16.2.25.pdf/0b498c8a-85aa-e089-32a2-18da59ee658c?t=1740988650613. ↩
- Swiss Criminal Code (SR 311.0), Art. 321 para. 1, bilingual edition, https://www.droit-bilingue.ch/de-en/3/31/311.0-321-389.html (lawyers, defence lawyers “sowie ihre Hilfspersonen”); Federal Act on the Free Movement of Lawyers (BGFA, SR 935.61), Art. 13 para. 1 (“Berufsgeheimnis”), https://www.droit-bilingue.ch/de-fr/9/93/935.61-13-17.html. ↩
- Swiss Bar Association, “Nutzung von Cloud-Diensten”, https://digital.sav-fsa.ch/en/digitale-kanzlei-nutzung-von-clouddiensten (accessed September 2026), summarising the opinion by Thouvenin, Schwarzenegger and others (ITSL, University of Zurich, 2019). ↩
For general information only. This article does not constitute legal or tax advice, does not replace advice on an individual case and creates no lawyer-client relationship. It reflects the legal position as of September 2025.