Chatbot technology is disrupting the ways in which companies interact and engage with their customers. As software has become more and more sophisticated and artificial intelligence and natural language processing more developed, the use of chatbots has significantly increased in the last few years and is expected to grow further. In this blog post, we will provide an overview of the legal considerations connected with the use of chatbots.
What Exactly Are Chatbots?
Chatbots are software programs that can simulate a conversation with a user in natural language, mainly through messaging applications, websites and mobile applications. Simple chatbots (so called clickbots) respond to keywords with answers that are pre-programmed into their systems. Smart chatbots are designed to understand and respond to a conversation in a natural, human-like manner. In order to do this, smart chatbots are equipped with artificial intelligence and machine learning. Through interactions with people and by accessing knowledge databases, smart chatbots become smarter and more personable after each conversation. This blog post will focus on smart chatbots.
Where Are Chatbots Used and Why?
Chatbots are used in a variety of different contexts, such as in online customer support, in e-commerce purchase processes, for online bookings, customer feedback or for marketing purposes. For example, by using chatbots in online customer support, companies can provide extensive and always-available customer assistance whilst at the same time saving costs. From a marketing perspective, companies can use chatbots to increase their customer engagement and monitor consumer data.
Chatbots are increasingly used in regulated industries, such as in financial or healthcare services. Banks have introduced chatbots to perform tasks that range from collecting basic customer information and checking an account balance to offering users suggestions on how to save money. Where chatbots are used in a regulated industry they must be programmed to comply with the applicable regulatory framework. For example, where a chatbot collects information of a bank customer, it must be ensured that the chatbot complies with banking secrecy laws.
Do Customers Have to Be Informed About the Use of Chatbots?
It is not always obvious for a chatbot user that he or she is talking to a piece of software rather than an actual person unless this is clearly indicated with a disclaimer on the website where the chatbot is used or by information through the chatbot itself. There is no express legal obligation in Switzerland to inform customers about the use of artificial intelligence. However, if customers are not informed that they are interacting with a chatbot, the Federal Act against Unfair Competition (UWG) could apply. This is especially relevant in situations where it is important for customers that they are interacting with an actual person. The UWG states that any conduct or business practice which is deceptive or otherwise contrary to the principle of good faith and which affects the relationship between competitors or between supplier and purchaser is unfair and unlawful. For example, a company could act deceptively within the meaning of the UWG if it operates a chatbot which provides financial advice to customers and describes itself as a knowledgeable person with the necessary experience without indicating that it is actually a chatbot. In this case, the customer would expect to be advised by a human being and not a chatbot.
What About Data Protection?
What if Chatbots Provide Wrong Information to a Customer?
If a chatbot provides wrong information or inaccurate advice to a customer, the customer could make a wrong decision and as a result thereof suffer a financial loss and even bodily harm. However, since chatbots are software programs which do not have their own legal personality, chatbots themselves cannot be liable to the customer. Instead, the company which operates the chatbot and makes it accessible to the customer could become liable for the financial loss based on contract or tort. In the event of bodily harm also the manufacturer of the chatbot might become liable based on the Product Liability Act if the bodily harm was caused due to a defect in the chatbot software. A liability based on contract requires that the financial loss (with or without bodily harm) was caused by a breach of a contractual obligation by the chatbot operator and that the chatbot operator acted with intent or negligence. The specific contract in question will have to be examined but the provision of wrong or inaccurate information through a chatbot most likely constitutes a breach of contract. The chatbot operator would act negligently, for example, if it failed to properly test the chatbot, failed to correct known or apparent errors or failed to ensure ongoing checks on the flawless functioning of the chatbot. Alternatively or if there is no contract between the chatbot operator and the customer, the chatbot oberator could become liable based on tort. For pure financial losses without any bodily harm a liability based on tort requires that the chatbot operator has breached a written or unwritten rule which aims to protect the financial assets of the customer.
What if a Chatbot Becomes Abusive?
There is a risk that chatbots which are provided with incorrect or incomplete training data or opened up to public access, could misbehave by providing abusive responses to a customer. For example, in 2016 Microsoft’s chatbot Tay started to tweet highly offensive and abusive content within a few hours of its release on Twitter. If a chatbot provides abusive or offensive responses to a customer, the personality rights of the customer could be unlawfully infringed and as a result the customer could bring a claim for damages and satisfaction against the chatbot operator. In addition to the infringement of customers’ personality rights, a misbehaving chatbot can negatively affect the company’s image and reputation.
Chatbots have many advantages for companies. In order to reduce the legal risks associated with them, companies should adequately inform their customers about the use of chatbots either by a disclaimer on their website or by information through the chatbot itself and have a privacy notice in place which informs customers about the data that is collected through the chatbot. Liability and reputational risks connected with the chatbot may be reduced by testing and reviewing the chatbot thoroughly through random conversations and by ensuring that the chatbot contains an intelligent censoring mechanism and a built-in oversight. It is further helpful for companies to have an internal policy in place which covers the permitted activities of the chatbot, the extent of information fed to it, data collection and processing, the maintenance of the chatbot and the monitoring of chatbot activities including triggers for human intervention.
 However, on 22 May 2019 the Organisation for Economic Co-Operation and Development (OECD), to which Switzerland is a member state, issued recommendations on the use of artificial intelligence in which it stated that a responsible approach to artificial intelligence involves informing data subjects about the interaction with artificial intelligence. Although the OECD recommendations are not legally binding for member states they carry moral force and it is expected that member states should try to implement them.