SAV-Wegleitung für den Umgang mit künstlicher Intelligenz
Schweizerischer Anwaltsverband / Fédération Suisse des Avocats (Swiss Bar Association)
last revised 2024-06-14last verified 2026-09-042 versionscountry page →
GuidanceIn forceVerification dutyConfidentialityClient consentDisclosurePermissive
map categories:Bar guidance on generative AIDuty to verify AI outputClient data in AI toolsClient consent and transparencyVendor due diligenceSupervision of AI workHorizontal AI statute
The Swiss Bar Association's guideline on AI, approved 14 June 2024 and updated 16 February 2025. It focuses on generative AI, sets a three-route test for putting client information into a tool, requires independent verification of output, addresses liability, copyright and notification duties, and explains the Swiss, EU and Council of Europe regulatory position.
- Applies to
- lawyers, firms
- Effective
- See version history
- Current version
- Approved by the SAV board · 14 Jun 2024
- Last verified
- 04 Sept 2026
What the document requires
- Before using any AI, clarify what happens to input data, who can access it and where it is stored, having regard to the Anwaltsgeheimnis, the DSG and the Standesregeln.
- Only three routes permit confidential input: an on-premise install where no data leaves the firm network; compliance with the SAV's IT-outsourcing and cloud rules; or an informed consent-and-waiver from the client.
- Output must be checked independently and critically; an AI cannot verify its own output. Failure sources named are hallucination, missing or dated training data, sycophancy and bias.
- The lawyer is liable for defective performance and cannot plead that the AI made the mistake; a liability limitation should be agreed with the client where lawful.
- A duty to disclose AI use may arise from the provider's terms, or where the client requires or legitimately expects personal performance.
- Switzerland will not transpose the EU AI Act; the Federal Council decided on 12 February 2025 to adopt the Council of Europe AI Convention and legislate sector-specifically, with an EJPD consultation draft due by the end of 2026.
Version history
| Version | Date | What changed | Source |
|---|---|---|---|
| Approved by the SAV boardcurrent | 2024-06-14 | Original guideline approved by the SAV Vorstand on 14 June 2024. | official |
| Updated version | 2025-02-16 | Updated to reflect the Federal Council's 12 February 2025 decisions on AI regulation and the BAKOM Auslegeordnung. | official |
Sources
Cite this record
SafeLegalAI Legal AI Regulation Tracker, "SAV-Wegleitung für den Umgang mit künstlicher Intelligenz" (Schweizerischer Anwaltsverband / Fédération Suisse des Avocats (Swiss Bar Association), Approved by the SAV board, 2024-06-14), safelegalai.com/regulation/other/ch-sav-wegleitung-ki (accessed 2026-09-04). Data: CC BY 4.0.
More official documents in Other
- Letter No. 112/WKMA.Y/HK2.1/IX/2026: draft judgment-writing guidance and templates for the new Criminal Codes · Mahkamah Agung
- AI in the advocate's practice: opportunities, tools and safe use [DI advokato praktikoje: galimybės, įrankiai ir saugus naudojimas] · Lietuvos advokatūra
- Supreme Court of Japan generative-AI demonstration for civil trials (FY2027 budget request) · Supreme Court of Japan
- Recommendation Guide on the Use of Artificial Intelligence for Lawyers [Avukatlar İçin Yapay Zekâ Kullanımı Tavsiye Rehberi] · Türkiye Barolar Birliği (TBB)
- Decision No. 1671/QD-TTg approving the National Strategy on Artificial Intelligence to 2030, vision to 2045 · Prime Minister of Vietnam
- Guideline on the Use of Artificial Intelligence for Indonesian Advocates [Pedoman Penggunaan Kecerdasan Buatan] · PERADI