Skip to content

AI agent liability in Switzerland: what binds you, who pays, when it is a crime

An AI agent is not a person. What it does binds, in principle, the individual or company using it. Bank, provider, crime: evidence decides.

In short. An AI agent has no legal personality under Swiss law. What it orders, signs or pays with your credentials, within the task you gave it, is in principle attributed to you, as if you had done it yourself. Three questions then need to be kept apart. Are you bound by the contract? Who bears the loss? Has a criminal offence been committed, and by whom?

With your bank, everything turns on whether the order came from a third party who took over your access, or from your own agent, tricked while doing what you had asked it to do. To our knowledge, no Swiss court has yet ruled on an AI agent. The outcome often depends on evidence, and evidence is prepared before the incident.

The procurement agent was supposed to order ten ink cartridges. It ordered a hundred. Or the treasury agent read a supplier invoice containing, in invisible text, an instruction to pay a different IBAN. It paid. The next day, the bank replies that the order was given with your credentials.

This is no longer hypothetical. Gartner’s 2026 survey of IT leaders found that 17% of organisations had already deployed AI agents and that more than 60% expected to do so within two years. In July 2026, the security firm Zscaler described two live campaigns hiding instructions in web pages to push AI agents into making payments. In its own test, 4 of the 26 models evaluated made the requested payment.

In Switzerland, according to a survey of 1,260 people aged 18 to 65 carried out for Visa and published on 15 October 2026, 70% have already used an AI assistant to prepare a purchase, but only 17% would let it pay on its own, and 66% make approving each payment themselves a condition of their trust. In July 2026, Visa announced the first purchases made by AI agents at merchants in Europe, with more than 30 card issuers, including Cornèrcard, Swisscard and Viseca.

Is an AI agent a person under Swiss law?

No. Software has no rights or obligations of its own. Under the Civil Code, civil rights belong to persons (art. 11 CC), and Swiss law contains no specific rules for AI agents. The Federal Supreme Court drew the consequence for software that carried out market making on its own. Lacking legal personality, the program was a mere tool, not a substitute within the meaning of art. 398(3) CO (Federal Supreme Court 4A_305/2021 of 2 November 2021, consid. 7.3.1). The Federal Administrative Court reasoned the same way in patent law, refusing to let an AI system be named as an inventor (Federal Administrative Court B-2532/2024 of 26 June 2025, consid. 4.9). To our knowledge, no published Swiss decision yet deals with an AI agent that contracts or pays.

Swiss legal writing draws the same conclusion, that such systems remain tools. Fabien Liégeois writes that the person using the system answers for its acts as for their own, and that, as the law stands, it can be neither an agent under arts. 32 et seq. CO nor an auxiliary under art. 101 CO (SZW/RSDA 2024, pp. 659-660).

In Europe, the German Federal Court of Justice set out the principle as early as 2012, in a case on an online booking system. It is not the computer system that makes or receives the declaration, but the person or company using it as a means of communication. The meaning of the declaration depends on how the human recipient could understand it in good faith, not on how the machine would process it (BGH, 16 October 2012, X ZR 37/12, para. 17).

The UNCITRAL Model Law on Automated Contracting, adopted in 2024, sets the same rule internationally. Unless the parties agreed on a procedure, an action carried out by an automated system is attributed to the person who uses the system for that purpose, even if the outcome was unexpected (art. 7). An optional provision covers the case where the other party knew, or could reasonably be expected to know, that the user did not expect the action (art. 8). The report submitted to the Federal Council by the Federal Office of Communications on 12 February 2025 leaves open whether Switzerland will take it over.

The only appellate ruling we know of on an AI agent as such is American. On 4 August 2026, the US Court of Appeals for the Ninth Circuit held that a browsing agent shopping on Amazon for its users is “a tool, not a person for statutory purposes”, and that it is the user who acts “with the help of” the agent (Amazon.com Services v. Perplexity AI, No. 26-1444). The court was careful to add that it was not creating a legal regime for agentic AI and was not deciding tort liability. The “tool” idea is a sound starting point. On its own, it does not settle what is attributed to you.

Are you bound by what your AI agent ordered?

In principle, yes, if you authorised it to contract and it stayed within that remit. Under the Swiss Code of Obligations (CO), a contract is formed when the parties have mutually and consistently expressed their intent, expressly or tacitly (art. 1 CO). Three situations follow different rules.

The agent did what you entrusted it with

You are bound. In our view, an order placed by an agent you launched with your own credentials, within the scope you left it, is attributed to you just as an order placed through an automated form would be. The seller may rely on the objective meaning of the order. Where the parties’ actual common intent cannot be established, the court looks at the meaning the other party could give the declaration in good faith, even if it does not match its author’s inner intent (art. 18(1) CO, BGE 144 III 93, consid. 5.2.3). If you let the agent choose the product or the supplier, a disappointing choice is not a legal mistake. It is your choice, delegated.

The agent did not express what you wanted

You asked for ten cartridges and the agent declared a hundred. Your intent was defined and was transmitted inaccurately. You may then rely on fundamental error, in particular where a party intended a different contract from the one it agreed to, or where the performance promised is significantly greater than intended (art. 24(1) nos. 1 and 3 CO). The Code also applies the rules on error by analogy where a party’s intent was inaccurately conveyed by a messenger or other intermediary (art. 27 CO). Applying this to an AI agent has not yet been tested in court, and it presupposes a specific intent to convey. In Germany, the Federal Court of Justice treated a price distorted by faulty software during transmission as an error in the declaration (BGH, 26 January 2005, VIII ZR 79/04). Two limits. A party whose error results from its own fault must compensate the other party, unless the latter knew or should have known of the error (art. 26(1) CO). And the contract is deemed ratified if you let one year pass from discovering the error without declaring that you will not uphold it (art. 31 CO). Write to the seller straight away.

Someone else took control

If a third party hijacked the agent or your access, it is not your intent being expressed. The seller may rely on apparent authority, but only on strict conditions. The Swiss Federal Supreme Court protects a third party acting in good faith on apparent authority where that appearance is attributable to the principal (art. 33(3) CO, BGE 120 II 197). The mere use of stolen credentials does not create it. Applying agency rules to software remains, in any event, an analogy the courts have not yet endorsed. As for the limits you had set, their weight depends on what the seller knew. A limit that was communicated, and received, counts for more than a setting only you could see.

Must the bank reimburse a payment made by your AI agent?

It depends on whom the order is attributable to. An order placed by a third party who took over your access is not yours. An order placed by your agent, which you tasked with this kind of payment, is in principle yours, even if the agent was tricked. This is no marginal issue. In 2025, as the year before, fraud was the most frequent cause of written complaints to the Swiss Banking Ombudsman, with 304 cases in the area of accounts, payments and cards, up 15%.

A third party took control

The Federal Supreme Court has set out the principle: when a bank executes the instructions of an unauthorised third party, or of a representative acting beyond their authority, it acts without a mandate from the client and has no claim for reimbursement (Federal Supreme Court 4A_81/2018, consid. 3). A client who claims their balance does not have to prove fault on the bank’s part. The bank must show that it performed the contract properly. If it shows that it acted on orders transmitted and verified as agreed, the client must then prove that a third party usurped their identity or means of communication (same judgment, consid. 5.3). This is why the agent’s logs matter, as they are often the only trace of the intrusion.

Two caveats can reverse the outcome. Bank terms often contain a so-called risk transfer clause, which places on the client the loss caused by an undetected unauthorised order. It does not apply in the event of the bank’s gross negligence (art. 100(1) CO by analogy, BGE 146 III 326, consid. 6.1). And the bank may set off a damages claim against a client who culpably contributed to the loss. The bank must prove breach, damage and causation, but the client’s fault is presumed, so it is for the client to show that no fault can be attributed to them (Federal Supreme Court 4A_119/2018, consid. 5.2). Poorly secured access can lead there.

Your agent was tricked

Here, the agent itself launched the payment, with your credentials, because a fake invoice or a hidden instruction misled it. In our view, what links this order to you is not merely the use of your access. It is the fact that you tasked the agent with paying this kind of invoice, using the authentication means agreed with the bank. The deception concerns the reason for the payment, not its authorisation, as in bank impersonation fraud, where the client approves the transfer personally. If the agent had no such authority, you are back in the previous case.

The loss then generally stays with you. Case law on bank verification concerns the authenticity of the order. A bank must carry out additional checks where there are serious signs of forgery, where the order does not relate to a transaction provided for by the contract or arising from practice, or where particular circumstances give rise to doubt (4A_81/2018, consid. 3). These rules do not create a general duty to check the invoice behind an authentic order. A claim against the bank remains conceivable if particular circumstances should have cast doubt on the order itself. It is rarely straightforward.

The European Union is taking a different route. The future Payment Services Regulation, politically agreed in November 2025 but not yet adopted, treats a transaction initiated or modified by a fraudster as unauthorised, and leaves the loss with a provider that failed to put appropriate fraud prevention in place. To our knowledge, Swiss law has no comparable statutory rule.

Is the AI provider liable for its tool’s mistakes?

First under its contract. A debtor who fails to perform, or performs defectively, is liable for the resulting damage unless it proves that it was not at fault (art. 97(1) CO). A clause excluding in advance liability for its own intent or gross negligence is void (art. 100(1) CO). Excluding liability for slight negligence is in principle possible, subject to the special cases of art. 100(2) CO and, towards consumers, to the ban on abusive standard terms (art. 8 of the Unfair Competition Act). For the acts of its auxiliaries, liability may be excluded wholly or partly by prior agreement (art. 101(2) CO). Caps and exclusions are in the provider’s terms. Read them before any exchange with the provider.

In Switzerland, the Product Liability Act

The Product Liability Act rarely helps. The producer is liable where a defective product causes death, personal injury or damage to property of a type intended and mainly used for private purposes (art. 1(1) PLA). A purely financial loss, such as a diverted payment, falls outside it. Whether stand-alone software is a “movable thing” under the Act (art. 3(1)(a) PLA) divides legal writers, and to our knowledge the Federal Supreme Court has not decided it. In its report to the Federal Council of 12 February 2025, the Federal Office of Communications notes a general need to modernise the Act. To our knowledge, no revision has been put forward.

In the EU, the new Product Liability Directive

The European Union has gone further. Directive (EU) 2024/2853 expressly includes software in the definition of product (art. 4(1)) and applies to products placed on the market or put into service after 9 December 2026 (art. 2(1)). It adds to the recoverable damage the destruction or corruption of data not used for professional purposes (art. 6(1)(c)). It also eases the victim’s burden of proof, since the court may order the defendant to disclose evidence and defectiveness and causation are presumed in certain cases (arts. 9 and 10). Purely financial loss remains outside its scope, as in Switzerland. Swiss companies selling an agent or software into the EU will need to factor this in.

The EU has, on the other hand, given up on a liability regime specific to AI. The Commission withdrew its proposed AI Liability Directive, with the withdrawal published in the Official Journal on 6 October 2025. Neither Switzerland nor the EU therefore has a special rule saying who answers for an AI agent.

Is your company liable for AI as it is for an employee, and for the agent it offers to its clients?

Not automatically, despite what is sometimes said. Employer liability covers damage caused by “its employees or other auxiliaries” (art. 55(1) CO), and a debtor’s liability for auxiliaries refers to “persons living in his household or employees” (art. 101(1) CO). The wording has people in mind. Whether they apply to software by analogy divides legal writers, and to our knowledge no court has accepted it. The Federal Supreme Court has even declined to treat a program as a substitute, for lack of legal personality (Federal Supreme Court 4A_305/2021, consid. 7.3.1).

The point matters. If the agent is treated as an auxiliary, the company is liable to its clients for the agent’s error even without any fault of its own (art. 101(1) CO). If the agent is merely a tool, the company remains bound by its contractual obligations according to their content, but it can escape liability by proving that it was not at fault (art. 97(1) CO), in how it selected, configured or supervised the agent. Its recourse against the provider is a separate matter, and caps accepted from the provider do not limit what it owes its own clients. Towards a third party with no contract, compensation requires damage caused unlawfully, intentionally or negligently (art. 41(1) CO). Swiss law has no general strict liability for software.

What the tool says, on the other hand, remains the company’s own conduct. On 12 May 2026, the Hamm Court of Appeal held, under unfair competition law, that a clinic’s chatbot that invented specialist titles was merely a technical means under the operator’s control, so that its answers were attributed to the operator (OLG Hamm, 4 UKl 3/25, appeal on points of law allowed). In Canada, an airline was held liable for wrong information given by its chatbot (Moffatt v. Air Canada, 2024 BCCRT 149).

The same reasoning applies to a company that offers an agent to its clients, to inform, advise, book or order. Towards them, it answers for what the agent says and does within the service promised, under the contract (art. 97(1) CO) and, absent a contract, under the conditions of art. 41 CO, without hiding behind the tool. Its terms of use may define what the agent is allowed to do and require human validation for binding acts, but no clause releases it from its own gross negligence (art. 100(1) CO).

For companies deploying AI agents in Switzerland. Define in writing what each agent may do, up to what amount and with whom. Agree payment limits and dual approval with your bank. Align what you promise your clients with what your AI provider actually guarantees, since its caps do not bind your clients. Keep logs for as long as legal retention rules and the limitation periods for likely claims require, and no longer than needed, since they may contain personal data. If you sell into the EU, factor in Directive (EU) 2024/2853 for products placed on the market after 9 December 2026.

An agent tricked into paying: is it a criminal offence?

Often, but the offence depends on the chain of events. Using an agent does not make you the perpetrator of an offence committed without your knowledge. Whoever tricked the agent, however, may be prosecuted.

Fraud under the Swiss Criminal Code (SCC) requires that a person was deceived by trickery, that this error led them to act against their own or a third party’s financial interests, and that the perpetrator intended unlawful enrichment (art. 146(1) SCC). Computer fraud was introduced for cases falling outside fraud because no person was deceived (BGE 129 IV 315, consid. 2.1). It covers anyone who, with the same intent, influences electronic data processing by using data incorrectly, incompletely or without authorisation, or by a similar process, and thereby causes, through an incorrect result, a transfer of assets to the detriment of another (art. 147(1) SCC). It carries a custodial sentence of up to five years or a monetary penalty, and six months to ten years if committed for commercial gain (art. 147(2) SCC). It is prosecuted ex officio, and only upon complaint where committed against relatives or members of the same household (art. 147(3) SCC).

In other words, two chains of events, two possible offences, each subject to all of its elements. If the agent paid on its own following a hidden instruction, computer fraud may be argued, provided it is shown that the instruction distorted the processing, caused the transfer and caused loss, and that its author sought unlawful enrichment. If a human approved the payment afterwards, fraud may be considered, provided that person was deceived by trickery about what they were approving and that this error caused the payment. A confirmation click is not enough. Either way, the criminal complaint must set out precisely what the perpetrator altered, what the agent read and did, and what the human saw.

How do you prove what the agent did?

Through its traces, which disappear quickly. The instructions given to the agent and their version at the time, the version of the agent itself, the content it actually read, its actions and the tools it called, human approvals, timestamps, and the bank’s authentication logs. Much of this sits with the provider or the platform. Ask them in writing, without delay, to preserve it, with a precise list.

Three legal levers exist. The right of access under the Federal Act on Data Protection (FADP) gives you the personal data processed and, where applicable, the existence of an automated individual decision and the logic behind it (art. 25(2)(b) and (f) FADP), as a rule within 30 days (art. 25(7) FADP). It guarantees neither access to all technical logs nor their preservation. In civil proceedings, a court may take evidence in advance where a risk to the evidence is shown to be likely (art. 158(1)(b) of the Civil Procedure Code). In criminal proceedings, the public prosecutor may seize items needed as evidence and order their surrender (arts. 263(1)(a) and 265 of the Criminal Procedure Code, CrimPC). A company may, however, refuse to surrender items where doing so would expose it to criminal or, under certain conditions, civil liability (art. 265(2)(c) CrimPC), which may concern the agent’s provider.

What should you do in the first hours?

Within the hour, suspend the agent and its access without deleting anything, and export what it did. The same day, dispute the payment with your bank in writing, ask for a recall of the funds and for the authentication logs. For a card payment, notify the issuer within the period set by its terms, block the card and ask in writing for a refund, without counting on an automatic chargeback. For a platform, open a claim and keep a record of it.

Within 48 hours, write to the seller that the transaction was not intended, ask the AI provider to preserve the logs, and file a criminal complaint if the agent was manipulated. If the bank refuses, the Swiss Banking Ombudsman offers a neutral mediation service free of charge.

Going forward, set written limits for your agents: a cap, a list of approved merchants, human approval above a threshold. Agree them with your bank where possible. These limits, and preserved logs, are what will make the difference when you have to prove your case.

If you are outside Switzerland

Each relationship has its own governing law, the contract with the seller, the one with your bank and the one with the AI provider. Which law applies depends on any choice of law, on the applicable conflict rules and, for consumers, on mandatory protections that a choice of law cannot remove. On the criminal side, an offence is deemed committed both where the perpetrator acted and where the result occurred (art. 8(1) SCC). Whether that is the case in Switzerland, for instance for a payment diverted to a Swiss account, has to be assessed on the facts. We work in English with clients and with foreign counsel.

What is Switzerland preparing?

On 12 February 2025, the Federal Council instructed the Federal Department of Justice and Police to submit a consultation draft by the end of 2026, with a view to ratifying the Council of Europe Convention on Artificial Intelligence, which Switzerland signed in March 2025. The priorities announced by the Federal Office of Justice are transparency, data protection, non-discrimination and supervision. Civil liability is not among them. The Federal Office of Communications report that prepared this decision considers Swiss liability law, with its general clauses, broadly fit for purpose. We will update this article when the draft is published.

Key deadlines at a glance

  • The same day, dispute the payment with your bank and ask for a recall. No statute sets this, but transfers are easier to recover when reported early.
  • The period set by your bank’s or card issuer’s terms for reporting a disputed transaction.
  • 30 days, as a rule, for a reply to a data access request under the FADP (art. 25(7) FADP).
  • One year from discovering the error to tell the seller you will not uphold the contract (art. 31 CO).
  • Three years from knowledge of the damage and of the liable person, and ten years at most from the harmful event, for a non-contractual damages claim. Where the damage results from a criminal act of the liable person, the claim does not become time-barred before the criminal limitation period expires (art. 60(1) and (2) CO).
  • Ten years for a contractual claim, unless otherwise provided, running from when the claim falls due (arts. 127 and 130(1) CO).
  • Three years from when you knew, or should have known, of the damage, the defect and the producer for a product liability claim, and ten years at most from when the product was put into circulation (arts. 9 and 10(1) PLA).

Frequently asked questions

My AI agent bought something I didn’t want. Am I liable under Swiss law?
In principle yes, if the agent stayed within the task you gave it, even if its choice disappoints you. If your intent was specific and the agent expressed it wrongly, you may rely on fundamental error, and the rules on error apply by analogy to intent wrongly conveyed by an intermediary (arts. 24 and 27 CO), which has not yet been tested for AI agents. Tell the seller in writing without delay that you will not uphold the contract.

Will my Swiss bank refund a payment made by my AI agent?
It depends on whom the order is attributable to. If a third party took over your access, a bank executing their instructions acts without your mandate, subject to any risk transfer clause and to your own possible fault. If your own agent, tasked with that kind of payment, was tricked, the order is in principle attributable to you and the loss stays with you, unless the bank failed in its own duties.

Who has to prove what against the bank?
The bank must first prove that it properly executed orders received and verified in the manner agreed with you. If it does, you must prove that a third party usurped your identity or access (Federal Supreme Court 4A_81/2018, consid. 5.3). The agent’s logs then become decisive.

Does a spending limit set in the agent protect me?
Not automatically. Its effect depends on what the seller or the bank knew and on what was agreed with them. A limit agreed with your bank or communicated to the seller carries more weight than a setting only you could see. In any case, it proves what you had authorised.

Is a company liable for its AI agent like it is for an employee?
Not automatically. Arts. 55 and 101 CO refer to employees and other auxiliaries, and to our knowledge no court has yet applied them to software. Towards its clients, a company remains bound by its contractual obligations, but it can escape liability by proving it was not at fault (art. 97(1) CO), unless the agent is treated as an auxiliary.

Someone tricked my AI agent into paying. Is that a crime in Switzerland?
It can be, provided all elements of the offence are met. If the agent paid on its own after a hidden instruction, computer fraud (art. 147 SCC) may be considered. If a human approved the payment after being deceived by trickery, fraud (art. 146 SCC). The complaint must describe precisely the chain of events, the loss and what the perpetrator was seeking.

Can the AI provider exclude all liability?
No. A clause excluding in advance liability for its own intent or gross negligence is void (art. 100(1) CO). Excluding liability for slight negligence and for auxiliaries is more widely accepted, subject to statutory limits and, towards consumers, to the ban on abusive standard terms (art. 8 Unfair Competition Act). The Product Liability Act does not cover purely financial loss.

Is there a specific AI liability law in Switzerland or the EU?
No. Neither Switzerland nor the EU has a special rule saying who answers for an AI agent, so the ordinary rules of contract, agency, tort and criminal law apply. Product liability law exists on both sides, the Product Liability Act in Switzerland and Directive (EU) 2024/2853 in the EU, which covers software placed on the market after 9 December 2026. Neither covers purely financial loss.

Legal basis and case law

Contract and error: arts. 1, 18, 24, 26, 27 and 31 CO, art. 11 CC. Agency and substitution: arts. 33(3) and 398(3) CO. Liability: arts. 41, 55, 97, 100 and 101 CO, art. 8 Unfair Competition Act. Limitation: arts. 60, 127 and 130 CO, arts. 9 and 10 PLA. Products: arts. 1 and 3 PLA. Offences: arts. 146 and 147 SCC. Evidence: art. 25 FADP, art. 158 Civil Procedure Code, arts. 263 and 265 CrimPC. EU law: Directive (EU) 2024/2853, arts. 2, 4, 6, 9 and 10. International: UNCITRAL Model Law on Automated Contracting (2024), arts. 7 and 8.

Legal writing: Fabien Liégeois, “Algorithmes contre algorithmes : de l’erreur humaine au dysfonctionnement de systèmes automatisés”, SZW/RSDA 2024, pp. 654-668.

Case law: BGE 120 II 197 (apparent authority), BGE 129 IV 315, consid. 2.1 (computer fraud), BGE 146 III 326, consid. 6.1 (risk transfer clause and bank’s gross negligence), Federal Supreme Court 4A_81/2018 of 29 May 2018, consid. 3 and 5.3 (order without mandate, duty to verify, burden of proof), Federal Supreme Court 4A_119/2018 of 7 January 2019, consid. 5.2 (bank’s claim against a negligent client), BGE 144 III 93, consid. 5.2.3 (principle of confidence), Federal Administrative Court B-2532/2024 of 26 June 2025, consid. 4.9 (AI system and inventorship), Federal Supreme Court 4A_305/2021 of 2 November 2021, consid. 7.3.1 (software without legal personality). German Federal Court of Justice (BGH), 16 October 2012, X ZR 37/12, para. 17 (declaration made through an automated system), BGH, 26 January 2005, VIII ZR 79/04 (price distorted by software), OLG Hamm, 12 May 2026, 4 UKl 3/25 (chatbot attributed to its operator). Moffatt v. Air Canada, 2024 BCCRT 149 (airline chatbot). US Court of Appeals for the Ninth Circuit, Amazon.com Services v. Perplexity AI, No. 26-1444, 4 August 2026.

Sources: Swiss Code of Obligations (Fedlex). Amazon v. Perplexity opinion (govinfo.gov). Directive (EU) 2024/2853 (EUR-Lex). BGH judgment X ZR 37/12. Federal Administrative Court, press release on B-2532/2024. European Parliament, withdrawal of the AI Liability Directive proposal. Visa, survey on AI in Swiss shopping (15 October 2026). Visa, first AI agent payments in Europe (2 July 2026). Gartner, Hype Cycle for Agentic AI (2026). SecurityWeek, prompt injection campaigns described by Zscaler (6 July 2026). Federal Office of Justice, artificial intelligence. Swiss Banking Ombudsman. Swiss Banking Ombudsman, annual report 2025. Federal Office of Communications, overview of AI regulation (12 February 2025). UNCITRAL Model Law on Automated Contracting. Liégeois, SZW/RSDA 2024 (UNIGE open archive). OLG Hamm, 4 UKl 3/25. European Parliament, Payment Services Regulation deal.

Our work on these matters

For legal support on digital tools, see our digital assets, Web3 and AI page, and for criminal complaints and proceedings, our criminal defence lawyer in Geneva page.

These cases are decided in the first days, between the bank, the tool provider and, where there was manipulation, the public prosecutor. A first meeting lets us review, with logs and contracts in hand, whether the transaction binds you, who can be pursued and what must be preserved. It does not prejudge the outcome. Sentinel Legal acts in payment fraud and economic crime, and Matthias Traussnig has spoken on blockchain-related fraud in the CAS Digital Finance Law at the University of Geneva. See also Hacked access keys: when your AI services bill explodes and Deepfakes in Switzerland: what criminal law covers. To discuss your situation: +41 22 512 76 00 or via the contact form.

Share this article LinkedIn
Matthias Traussnig
À propos de l'auteur Matthias Traussnig Avocat Barreau de Genève, Suisse

Me Matthias Traussnig exerce en droit pénal, en contentieux, en droit des nouvelles technologies et en droit des affaires (litiges pénaux et civils, Web3, finance numérique, régulation des crypto-actifs et cybersécurité). Il a exercé pendant trois ans au Ministère public du canton de Genève et siège comme juge assesseur à la Chambre pénale d'appel et de révision (CPAR) de la Cour de justice de Genève. Après une expérience en Big Four, il a été Senior Manager dans un cabinet actif dans les nouvelles technologies, aux côtés de nombreux projets blockchain et d'acteurs de la finance numérique. Membre de l'Ordre des avocats de Genève et de la Fédération Suisse des Avocats, il est intervenu dans le CAS Digital Finance Law de l'Université de Genève sur les fraudes liées à la blockchain. Il est le fondateur d'arnaques-suisse.ch.

Expertise Droit pénal, contentieux, nouvelles technologies et Web3, droit des affaires Languages Français, anglais, allemand Tous les articles de Matthias Traussnig →

Your situation deserves a careful reading.

Our team is available to review your case in strict confidence.