1 Legal and enforcement framework
1.1 In broad terms, which legislative and regulatory provisions govern AI in your jurisdiction?
Currently, there are no specific regulations or statutes governing artificial intelligence (AI) in Switzerland. However, in a report entitled "Défis de l'Intelligence Artificielle" ("The Challenges of Artificial Intelligence"), dated 13 December 2019, the Swiss Federal Council – the executive authority of Switzerland, comprised of seven ministers elected every four years by the two Swiss parliamentary chambers – concluded that existing contract, labour, tort and data protection legislation could encompass potential issues arising from the use of AI without requiring specific adaptations.
In its report, the Swiss Federal Council concluded that legislation which is too specific can quickly become outdated given advancements in new technology, and may thus hinder the development of AI in Switzerland.
Instead, Switzerland relies heavily on pre-existing legislation without specific reference to AI, privileging laws that are formulated in a technologically neutral way, meaning that AI-specific issues can be integrated within the scope of Swiss law without having to amend it.
However, Switzerland's preference for technology-neutral legislation does not mean that specific and precise legislative provisions will not be implemented in the future, to help regulate emerging issues relating to the application of AI.
1.2 How is established or ‘background' law evolving to cover AI in your jurisdiction?
As set out in question 1.1, there are no specific regulations governing AI in Switzerland. However, certain established laws are evolving to cover AI – in particular, the Swiss data protection regime, which comprises the Federal Data Protection Act (DPA) and the Federal Ordinance to the Federal Act on Data Protection (DPO). The Swiss Data Protection and Information Commissioner ensures compliance with the data protection regime and regularly publishes guidelines to address more specific issues such as the processing of medical or employee-related data.
The DPA is currently being revised in order to comply with the General Data Protection Regulation (2016/679) and address issues arising from the use of AI.
Based on the 2017 DPA revision project, these amendments include the following:
- Increased transparency when processing data: Private actors will have to inform data subjects of the collection and processing of their data; and
- An encouragement to self-regulate: Companies can establish codes of conduct on data processing and submit them to the Swiss Data Protection and Information Commissioner. If approved by the commissioner, they will then apply to the company. Given the complexities of data processing and AI, allowing actors in this field to self-regulate will help to provide tailored and adequate regulation. In cases where a decision which has legal or significant consequences for a person is taken exclusively via automatically processed data, the data subject may request that the decision be reviewed by a natural person.
1.3 Is there a general duty in your jurisdiction to take reasonable care (like the tort of negligence in the United Kingdom) when using AI?
As explained in question 1.1, established Swiss law governs this field. As such, the Swiss law of tort, as laid out in the Swiss Code of Obligations, governs the use of AI. General obligations to take reasonable care can be found in Articles 41 and 97 of the Swiss Code of Obligations. Article 41 applies to extra-contractual liability, while Article 97 applies to contractual liability. Both provisions include a duty of care.
However, there is no singular definition of the duty of care. This is rather based on how a diligent person would have behaved in similar circumstances. The duty of care is therefore analysed on a case-by-case basis. If the behaviour strays from how a diligent person would have behaved in similar circumstances, the behaviour will be considered tortious in case of Article 41 or faulty in case of Article 97.
1.4 For robots and other mobile AI, is the general law (eg, in the United Kingdom, the torts of nuisance and ‘escape' and (statutory) strict liability for animals) applicable by analogy in your jurisdiction?
See question 1.1.
Robots and mobile AI devices are subject to general Swiss tort rules. Apart from the tort rules mentioned in question 1.3, other laws regulate specific tort-related issues, such as the Swiss Product Liability Act. In order for this law to apply, there must be a product. Under the Swiss Product Liability Act, a ‘product' is any moveable object, as well as electricity. Case law considers software to be a product even though it often is not a moveable object (ie, not on a CD or USB stick, but downloaded from an online resource). A software producer can be liable for any damages caused by faulty software. For mobile AI devices – which are a combination of software and hardware – the fact that software is incorporated in the hardware means that they constitute a ‘product' under Article 3 of the Swiss Product Liability Act. The software producer can therefore be liable for any damages caused by the faulty software.
If the mobile AI at issue causes damage to someone due to a fault, the producer of the mobile AI will be liable pursuant to Article 1 of the Swiss Product Liability Act and Articles 16 to 18 of the Product Safety Act.
On the other hand, if the cause of the damage lies in the product's hardware and the software is produced by a separate entity, the manufacturer of the hardware, rather than the software developer, will be liable.
1.5 Do any special regimes apply in specific areas?
As set out in question 1.1, there are no specific regulations or statutes governing AI in Switzerland. However, given the potential applications of AI, numerous laws will ultimately regulate the use of this technology:
- The Swiss Data Protection Law is currently being revised to better incorporate AI (see question 1.2).
- Protection against unfair competition is assured by the Swiss Federal Act against Unfair Competition and the Swiss Cartel Act. Under Article 5, para 3 lit a of the Swiss Cartel Act, agreements between companies which indirectly or directly fix prices are prohibited. Agreements between competitors to fix prices using AI are prohibited under the Swiss Cartel Act.
- A similar issue is the use of AI in financial trading. Under Article 143 of the Financial Market Infrastructure Act, it is prohibited to use algorithms to display false or potentially misleading information on supply and demand of financial products. Article 31 of the Swiss Financial Market Infrastructure Ordinance clearly states that institutions using algorithms for trading purposes must record all orders given by these systems. They must also take specific precautions to avoid disruptions to the financial market.
- Swiss discrimination law, as set out in Article 328 of the Swiss Code of Obligations, may also potentially apply to issues arising from the use of AI. This is particularly relevant in the hiring process, where AI-powered solutions are increasingly used to help companies with their HR needs (see question 6.1).
1.6 Do any bilateral or multilateral instruments have relevance in the AI context?
Switzerland joined the World Intellectual Property Organization (WIPO) in 1970 and is a participant in the WIPO Conversation on Intellectual Property (IP) and Artificial Intelligence (AI). The goal of the WIPO Conversation is to provide member states with an opportunity to discuss different topics relating to AI and, more specifically, to discuss how AI is challenging the current global IP framework and how to address the challenges presented by this new technology.
Moreover, Switzerland is in the process of ratifying the amendment to the Council of Europe's Convention 108, also known as the Convention for the Protection of Individuals with Regard to the Processing of Personal Data. This amendment addresses issues of AI, such as the automatic processing of data, and led to the amendment of the Swiss Data Protection Law in order to comply with Switzerland's international engagements. (See question 1.2 for amendments of particular relevance for AI in Switzerland.)
1.7 Which bodies are responsible for enforcing the applicable laws and regulations? What powers do they have?
The institutions responsible for enforcing Swiss law and regulations are the Swiss national courts. However, as Switzerland is a federation, the cantonal courts are competent at first instance, while the Swiss Supreme Court is the highest judicial authority in Switzerland. As such, the Swiss Supreme Court is the final instance for all appeals against decisions of the highest cantonal courts, the Federal Criminal Court, the Federal Administrative Court and the Federal Patent Court. The Swiss Supreme Court ensures that Swiss federal law is correctly applied in individual cases, and that the rights of citizens enshrined in the Swiss Constitution are protected.
The Federal Patent Court will decide whether a patent for a technical innovation is legally valid and whether patent rights have been infringed by a product or process. It will further hear disputes over who holds a given patent or how a patent may be licensed.
With regard to data protection, the Swiss Data Protection and Information Commissioner is responsible for ensuring compliance with the data protection laws. The commissioner may refer the case for a formal decision to:
- the competent department of the Swiss Federal Chancellery (in case of recommendations to federal bodies); or
- the Swiss Federal Administrative Court (in case of recommendations to private persons).
Appeals may be filed with the Swiss Supreme Court.
1.8 What is the general regulatory approach to AI in your jurisdiction?
See question 1.1.
2 AI market
2.1 Which AI applications have become most embedded in your jurisdiction?
Several different AI applications are being developed in Switzerland. Their primary focus is on reducing the costs of running operations for various types of industries through automatisation. When it comes to full product development for market availability, as opposed to the proof of concept phase, the Swiss market has adopted a rather cautious approach. Only a small number of businesses in Switzerland have been able to offer AI products to the market. The biggest challenge for companies seeking to produce AI products relates to scaling an AI model for production and market distribution.
The following AI applications are either in development or in use in Switzerland:
- The Swiss Army is running a research programme called Technology Foresight in its Armasuisse Science and Technology Department, which specialises in identifying and developing new technologies that disrupt national defence.
- Gamaya SA, a spin-off of the Polytechnical Federal Institute of Lausanne, uses AI and hyperspectral imaging to provide agronomy solutions to farmers.
- Squirro AG is an AI engine that provides companies with recommendations for better business outcomes by compiling complex information to help users increase revenue.
- SCAN (the Support Centre for Advanced Neuroimaging) in Bern is working towards machine learning-oriented solutions to predict strokes and epileptic disorders by processing large amounts of radiographs from patients.
2.2 What AI-based products and services are primarily offered?
See question 2.1.
2.3 How are AI companies generally structured?
Switzerland has the highest number of AI companies per capita in the world. An interview conducted with Axel Uran of Swiss-based AI company Visium SA and available market data indicate that most AI companies in Switzerland are at the start-up level. Many of these companies are structured as a stock corporation, also known in Switzerland as a Société Anonyme (SA) or Aktiengesellschaft. Companies that specialise in AI – such as Sophia Genetics, Nnaisense and Recapp – are all incorporated as such.
This specific structure is particularly useful as shareholder liability for company losses is limited to their share contribution. The company is therefore solely responsible for its debts. The Swiss SA is also attractive to future shareholders because the number of shareholders is potentially unlimited. This is not the case with a limited liability company, also known as a Société à Responsabilité Limitée or Gesesllschat mit beschränkte Haftung, where the maximum permitted shareholders is capped at 100.
2.4 How are AI companies generally financed?
The financing of AI companies depends on the type of company. For example, AI start-ups in Switzerland are financed primarily by venture capital funds and through funds available through Swiss innovation funding programmes and the European Union. Larger established companies that are seeking to grow in the AI field primarily fund AI initiatives through budget allocation for research, development and innovation.
2.5 To what extent is the state involved in the uptake and development of AI?
See question 8.1. Switzerland also promotes and encourages innovation in general, and the development of AI in particular. For example, Swiss federal agencies are dedicated to the active development and promotion of innovation through the deployment of AI technology, including:
- Innosuisse – the Swiss Agency of Innovation Promotion (www.innosuisse.ch/inno/de/home.html); and
- the State Secretariat for Education, Research and Innovation (www.sbfi.admin.ch/sbfi/de/home.html)
3 Sectoral perspectives
3.1 How is AI currently treated in the following sectors from a regulatory perspective in your jurisdiction and what specific legal issues are associated with each: (a) Healthcare; (b) Security and defence; (c) Autonomous vehicles; (d) Manufacturing; (e) Agriculture; (f) Professional services; (g) Public sector; and (h) Other?
As stated in question 1.1, there are no specific regulations relating to AI. Instead, general Swiss law is applied in the AI context. The primary regulatory issue in the healthcare industry in Switzerland involves the treatment of patient medical data and how it is processed (see question 1.2 for a general discussion of the Swiss data protection regime).
Because medical data is considered sensitive according to Article, 3 lit c (2) of the Federal Data Protection Act (DPA), express consent must be given prior to its processing. Where the data is anonymised, however, one must ensure that there is no way to identify the data subject. Failure to do so can lead to profiling, according to Article 3, lit d of the DPA, and is therefore unlawful according to Article 12 of the DPA. However, if the data has been properly anonymised, it is not subject to the DPA.
(b) Security and defence
As stated in question 1.1, there are no specific regulations relating to AI. Instead, general Swiss law applies in the AI context.
(c) Autonomous vehicles
As stated in question 1.1, there are no specific regulations relating to AI. Instead, general Swiss law applies in the AI context.
In Switzerland, autonomous vehicles (AVs) are currently being tested, albeit in very specific and limited situations. Legal issues relevant to any vehicle will equally apply to AVs.
By way of analogy, if an AV were to cause damage, the general regime set out in the Federal Law on Road Traffic would apply. The owner or the driver would be liable for the damages caused based on Article 58 of the Federal Law on Road Traffic. In that case, the basis for liability would lie solely in the result, as anyone who benefits from an AV is considered to have accepted the risks associated therewith. It is interesting to note that the owner would be liable regardless of any fault committed by him or her.
As stated in question 1.1, there are no specific regulations relating to AI. Instead, general Swiss law applies in the AI context.
In the context of manufacturing, the general rules that apply are contractual and are regulated by Article 97 of the Swiss Code of Obligations. Given that manufacturing is still primarily undertaken by humans, issues in the context of manufacturing may arise regarding the employer's liability for loss or damage caused by its employees or ancillary workers when executing their work. Such obligations are defined in Article 55 of the Swiss Code of Obligations.
An employer will therefore be liable for damages caused by its employees while performing their work, unless it can prove that all necessary precautions were taken to prevent the damaging event from occurring.
The employer, under Article 55, paragraph 2, has a right of recourse against the employee if his or her behaviour was negligent.
Agricultural methods are being transformed by the development of increasingly advanced ‘smart farming' technologies. These aim to make farms more efficient, optimise processes, increase yields and minimise their impact on the environment. This is a revolution with many potential advantages, which Agroscope – the centre of competence of the Swiss Confederation for Agricultural Research – is studying with the aim of enhancing the competitiveness of Swiss farming.
Research is also focusing on the use of ‘big data' in agriculture. Collected by intelligent objects equipped with sensors and connected to each other, the data analysed by algorithms aims to facilitate the efficient use of resources, which are also optimally adapted for purpose. For example, data on a cow collected by a robotic milker enables the robot supplying the fodder to adapt this supply to the needs of the animal in question. This is all done by algorithms; there is no need for the farmer to intervene. Another example is the Agrométéo web platform, which monitors everything from the evolution of mildew to insect populations and the phenological development of wheat. All of the information needed for plant protection is collected on a single site for farmers. The different applications assist in decision making to manage plant protection issues relating to large-scale crops, viticulture and arboriculture.
(f) Professional services
Professional services companies are increasingly using AI to tackle one of the biggest problems when dealing with clients: document acceptance, maintenance and management. Each type of data presents its own unique problems, which advancements in optical character recognition (OCR) technology are now helping to solve. OCR transforms data of all kinds into readable, useable electronic documents. For instance, advanced OCR technology assists lawyers in sifting through large discovery sets; helps medical practices to upload patient data into their systems; and serves any number of other uses. Once documents have been processed by OCR technology, they become data searchable, sortable and malleable.
Financial services firms are also incorporating AI into their strategies to drive operating and cost efficiencies, as well as critical business transformation programmes. They use AI solutions to design tailored services and products that better suit customers' needs, and to determine customers' individual risk profiles more effectively.
4 Data protection and cybersecurity
4.1 What is the applicable data protection regime in your jurisdiction and what specific implications does this have for AI companies and applications?
See question 1.2.
4.2 What is the applicable cybersecurity regime in your jurisdiction and what specific implications does this have for AI companies and applications?
The primary issue with AI and cybersecurity is the potential breach of personal data processing rules. Two Swiss laws are relevant here:
- Article 7 of the Federal Data Protection Act (DPA) and Articles 8 to 12 of the Federal Ordinance to the Federal Act on Data Protection (DPO) set out the general security requirements applicable to the processing of personal data; and
- Article 43 of the Federal Telecommunications Act sets out the principle of telecommunications secrecy.
Pursuant to the Swiss data protection regime, to prevent a breach, actors which use automatic data processing must implement measures which are "suitable" to achieve data security goals (Article 9, para 1 of the DPO). There is no specification as to what constitutes "suitable" security measures. Under the revised DPA, data controllers will have to report data breaches to the Swiss Data Protection and Information Commissioner if the breach may impact on a person's personality or fundamental rights. Where large numbers of data subjects are affected by the breach, the data controller may have a duty to publicly report the breach based on the principles of transparency and good faith.
Under the Federal Telecommunications Act, a provider of telecommunications services cannot disclose any information relating to communications between subscribers. As data transfers constitute telecommunications services, internet service providers must keep the communications of their subscribers confidential.
The 2020 Ordinance on Protection against Cyber-Risks in the Federal Administration aims to combat cyber crime. This ordinance created the National Cyber Security Centre, which has become the single point of contact for all issues relating to cybersecurity in Switzerland.
5.1 What specific challenges or concerns does the development and uptake of AI present from a competition perspective? How are these being addressed?
Fair competition is regulated by the Competition Commission and the Swiss Cartel Act. From a competition perspective, the increasing use of AI raises concerns due to the centralisation of large amounts of data in the hands of the few AI companies that process such data. This situation could lead to collusion between AI companies and/or a situation where an AI company attains an abusively dominant position, both of which are illegal under the Swiss Cartel Act as follows:
- Under Article 5, collusion between actors in a given market is unlawful if they either eliminate or, without justification, significantly restrict competition. Such collusion is unlawful only if it is intentional; and
- Under Article 7, any abuse of a dominant position is unlawful. More particularly, dominant actors are prohibited from hindering other companies from starting or continuing to compete, and from disadvantaging trading partners.
Collusion: No sanctions would apply if AI-infused devices were to accidentally collude. However, if the algorithm-driven behaviours were intended to collude so as to restrict competition, such collusion would be considered unlawful under Article 5.
Abusively dominant position: As AI companies can store, collect and process large amounts of data, antitrust concerns must be taken into account. AI companies that have access to large amounts of data might be placed at a considerable advantage vis-à-vis companies with limited access to data. Thus, determining what constitutes an abusively dominant position will be critical in determining whether AI companies can be held liable for anti-competitive practices under Swiss law.
6.1 What specific challenges or concerns does the development and uptake of AI present from an employment perspective? How are these being addressed?
The recruitment process and the processing of employee data by employers will become increasingly challenging with the rise of AI.
Discrimination – for example, on the basis of race, gender, national origin, religion, disability or age (some of the protected categories in Switzerland) – may arise when AI is integrated into the recruitment process. For example, if an AI algorithm used for recruitment is trained only with data using categories that are protected from discrimination under Swiss law, this would lead to ‘machine bias' and could amount to a violation of Swiss law, such as labour law, the Gender Equality Act or the Disability Discrimination Act.
For example, Swiss labour law, through Article 328 of the Code of Obligations, prohibits any unjustified discrimination in the recruitment process. AI hiring solutions thus cannot be programmed with a bias towards legislatively protected categories under Swiss law.
In terms of the processing of employee data, Article 328b of the Code of Obligations provides that employers may handle employee data only to the extent that this relates to the employee's suitability for employment. Thus, using AI to process employee data for reasons unrelated to the employee's employment contract – such as race – could violate Swiss law.
Plaintiffs that bring actions based on discrimination may request the full range of remedies available under the various Swiss statutes that prohibit discrimination based on protected categories, including injunctive relief to cease the discriminatory acts and monetary damages (both direct and indirect).
7 Data manipulation and integrity
7.1 What specific challenges or concerns does the development and uptake of AI present with regard to data manipulation and integrity? How are they being addressed?
One of the challenges posed by AI is its lack of transparency. For example, it is difficult to assess how an AI device reached its conclusion. The Data Protection Act discussed in question 1.2 is currently being revised to address this issue, among others.
8 AI best practice
8.1 There is currently a surfeit of ‘best practice' guidance on AI at the national and international level. As a practical matter, are there one or more particular AI best practice approaches that are widely adopted in your jurisdiction? If so, what are they?
One of the objectives of Switzerland's Digital Strategy is to maintain and develop the country's position as a global leader in AI. The Swiss Digital Strategy is a set of measures taken by the Swiss government to address emerging technology-related issues. These measures focus on many fields, including education, research, innovation, security, energy, cyber-administration, the economy, data, digital content, AI, culture and international involvement. The goal is for Switzerland to take advantage of the opportunities afforded by new technologies and to remain a worldwide leader in the field.
On 11 September 2020, the Federal Council, set out seven guiding principles for its 2020 Digital Strategy, as follows:
- A human-centred approach: Dignity, wellbeing and the public interest should be priorities when developing and using AI. Special attention should be given to the quality and balance of the data used, to prevent bias.
- Proper conditions to develop and use AI: Swiss law should remain technologically neutral, and education, research and innovation should remain a priority.
- Transparency, traceability and explainability: The processes and conclusions reached by an AI application should be identifiable, verifiable and traceable.
- Responsibility: One should establish who is responsible when using AI. Responsibility cannot be delegated to machines.
- Security: From the development stage onwards, AI systems should be safe, robust and resilient, so that they cannot be manipulated and lead to undesirable consequences.
- Active participation in AI governance: Players in the field should participate in global discussion.
- Implication of all relevant national and international actors: All stakeholders should be included in the AI governance process.
8.2 What are the top seven things that well-crafted AI best practices should address in your jurisdiction?
Building on question 8.1, a well-crafted AI best practice policy should address the following:
- fairness and lack of bias (vis-à-vis protected categories);
- human agency (including oversight);
- privacy and security;
- follow-on research; and
- communication of limitations.
8.3 As AI becomes ubiquitous, what are your top tips to ensure that AI best practice is practical, manageable, proportionate and followed in the organisation?
As for any new fields of interest and importance, the establishment of specialised AI management committees or advisory boards as a mechanism to monitor and assist organisations in developing appropriate responses to AI business, ethical and legal obligations will be crucial. As the saying goes, "What gets measured, gets managed."
9 Other legal issues
9.1 What risks does the use of AI present from a contractual perspective? How can these be mitigated?
The sale and use of devices that incorporate AI technologies present accrued risks in case of damages from a contractual perspective, due to the challenge of distinguishing who is liable for what.
AI may make a mistake by itself. Because a machine cannot act wilfully or negligently under Swiss law, responsibility must be appropriately allocated for any damages caused by the machine. The producer will therefore be liable for damages caused by a faulty device that incorporates AI technology. If the software provider is a different entity from that which manufactured the end product, it will be liable for the damages caused. If both the software and hardware are not faulty, the damages caused by the AI will be attributed to the person using the technology that has caused harm to a third party, regardless of fault. The person benefiting from the use of AI will thus assume the risks which are associated with it.
Because of these risks, sellers and users of devices that incorporate AI technology might consider shifting some of the risk to an insurer (see question 9.2).
9.2 What risks does the use of AI present from a liability perspective? How can these be mitigated?
There are two regimes of liability under Swiss law: contractual and non-contractual. Article 97 of the Swiss Code of Obligations regulates contractual liability, while Article 41 generally regulates non-contractual liability.
Contractual liability: Many AI services are governed by contracts. Article 97 provides that if an obligor fails to execute its obligation, then compensation is owed for the damages that result from this failure, unless it can prove that it was not at fault. The obvious contractual risk involved with the use of AI is that:
- the parties fail to apportion liability in the contract; or
- the selected apportionment violates Swiss law.
Thus, while the use of contractual provisions can help to mitigate the risks specific to the use of AI by setting out clear standards and specifications, such provisions must comply with Swiss law.
Non-contractual liability: Under Article 41, a person that causes loss or damage is obliged to provide compensation therefor. The burden of proof for any such loss or damage lies with the injured party. If the use of AI causes damage in Switzerland, the challenge is to distinguish whether the damage was caused by a faulty product (ie, mistakes that the AI made on its own) or by wilful programming. It may be difficult to assert which element of the product was faulty. However, this assertion is essential, as it defines who is liable in case of damage caused by the AI. One must therefore establish whether the damage is due to hardware, software or a combination of the two (see question 1.4).
9.3 What risks does the use of AI present with regard to potential bias and discrimination? How can these be mitigated?
See question 6.1.
10.1 How is innovation in the AI space protected in your jurisdiction?
According to Article 2, para 3 of the Swiss Copyright Act, software can be protected through copyright. An algorithm that generates AI can thus be subject to copyright; however, the algorithm must be integrated in software and not simply be a parameter. Such programs are protected for 50 years after the death of their creator. Creations made by AI cannot be subject to copyright, because they are not a creation of the human mind.
In order to be patentable, there must be a process that generates a result. It is this process or chain of actions which is patented, rather than the final result. An actual algorithm can therefore be patented; however, the results generated by AI are excluded from patent protection, because software lacks the legal capacity to be an inventor.
The Federal Act on the Protection of Designs does not address the issue of creations generated by AI. It is thus a complicated issue to determine who the designer is when dealing with an AI-generated creation. Is the designer the AI's producer, the person providing the training data sets or the end user? Because the creations of AI cannot be protected by law, there must be contractual provisions that clearly establish who holds the rights to the AI's creations.
10.2 How is innovation in the AI space incentivised in your jurisdiction?
The Swiss government plays an important role in incentivising innovation in AI, in relation to both funding and development. As mentioned in question 8.1, Switzerland's Digital Strategy aims to maintain and develop the country's position as a global AI leader. Innovation is incentivised by creating an optimal environment for the development of AI – in particular, in the following ways.
First, as set out in question 1.1, Switzerland does not overly regulate AI. Instead, a technology-neutral policy has been adopted. Given the absence of regulatory obstacles and Switzerland's beneficial taxation system, investment in AI has thrived.
Second, Switzerland encourages innovation through various research institutes and AI laboratories in order to generate more traction for the industry. The following state-funded institutions are some of Europe's leading technical institutes and attract some of the best minds in AI:
- the Federal Institute of Technology Zurich;
- the Federal Institute of Technology Lausanne; and
- the Dalle Molle Institute for Artificial Intelligence, Lugano.
In addition to these state-funded institutions, Switzerland is home to numerous independent centres, such as:
- the IDIAP Research Institute in Martigny;
- the Swiss Center for Electronics and Microtechnology; and
- the Swiss Academy of Engineering Sciences.
Third, Switzerland is a training ground for companies seeking to hire individuals with deep AI knowledge. Multinational tech leaders such as Google (Zurich), Hewlett Packard Enterprise (Geneva) and IBM (Zurich) conduct their AI research in Switzerland.
As a result, Switzerland has created an environment which should enable it to become one of the most important centres for AI technology in the world.
11 Talent acquisition
11.1 What is the applicable employment regime in your jurisdiction and what specific implications does this have for AI companies?
The main sources of employment law in Switzerland are the Code of Obligations, the Labour Act and the terms agreed in the contract of employment. In some industries, mandatory collective bargaining agreements are applicable. Additional legislation sets out specific rights and obligations, such as the Gender Equality Act, the Data Protection Act and the Merger Act.
As noted in question 6.1, the recruitment process will become increasingly challenging with the rise of AI. Talent acquisition is one area of application for AI technologies, which aim to automate recruiting. Employers that use AI tools must thus ensure that they do not violate privacy rights or the prohibition of unjustified discrimination as set forth in Article 328 of the Swiss Code of Obligations. For example, if an AI algorithm used for recruitment is trained only with data using categories that are protected from discrimination under Swiss law, this would lead to ‘machine bias' and could amount to a violation of Swiss law, such as labour law, the Gender Equality Act or the Disability Discrimination Act.
11.2 How can AI companies attract specialist talent from overseas where necessary?
Switzerland attracts an abundance of specialist talent, for numerous reasons. Switzerland is world renowned in fields such as training, research and innovation, as well as for its industrial and service sector companies. Swiss AI companies regularly score top marks for innovation, enhancing the country's attractiveness as an economic hub. Stable economic and political conditions are further factors that attract specialist talent from overseas.
According to a recent study by the International Institute for Management Development (IMD) in Lausanne: "the outbreak of the coronavirus pandemic has not dented Switzerland's potential to attract talent on a global scale… The training system, quality of life and high salaries remain the country's main strengths… Robust international trade fuels its strong economic performance, whilst its scientific infrastructure and health and education systems show steadfast displays."
Switzerland has a dual system for the admission of foreign workers. Workers from the European Union and from European Free Trade Area (EFTA) states have easier access to the Swiss employment market, thanks to the Agreement on the Free Movement of Persons; whereas only a limited number of workers are admitted from non-EU/EFTA states – primarily management level employees, specialists and other qualified employees.
12 Trends and predictions
12.1 How would you describe the current AI landscape and prevailing trends in your jurisdiction? Are any new developments anticipated in the next 12 months, including any proposed legislative reforms?
The COVID-19 crisis has introduced many unknowns and complicated trend predictions. The AI landscape in Switzerland is described in question 1.1. However, given Switzerland's dedication to maintaining its position as one of the most innovative countries in the world, particularly in the AI space, the following trends are likely to remain true into 2021 and beyond:
- The research and use of AI will continue to increase in Switzerland.
- Switzerland will continue to fund research programmes on the effective and appropriate use of AI.
- Switzerland will continue to be viewed as a training ground for AI talent, given the numerous universities and research institutes in Switzerland that focus on AI.
- These Swiss universities and research institutes will continue to provide a highly skilled pool of professionals to the many multinationals that are establishing research facilities in Switzerland. Indeed, these universities and research institutes are likely to forge strong partnerships with the growing number of established businesses in the AI space.
13 Tips and traps
13.1 What are your top tips for AI companies seeking to enter your jurisdiction and what potential sticking points would you highlight?
AI is one of the fastest-growing industries in Switzerland, which affords AI companies significant potential to start their own businesses in this field. AI companies seeking to commence activity in Switzerland should seek to collaborate with one of the key research institutes, such as the Federal Institutes of Technology in Zurich and Lausanne, or the Dalle Molle Institute for Artificial Intelligence in Lugano. Further, Swiss banks and telecommunications, insurance, health tech and pharmaceutical companies are continuously investing in AI and are thus partners which should be prioritised.
Although Switzerland ranks among the world's leading economies, it still lacks a well-established start-up ecosystem. Start-ups in Switzerland lack capital during their growth phase. For instance, a potential concern for AI newcomers in Switzerland might be the incorporation of a Swiss stock corporation, which requires a capital of CHF 100,000. However, the recently launched Swiss Entrepreneurs Foundation and the Swiss Entrepreneurs Fund may provide relief.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.