Artificial-intelligence systems that are available to the public are used not only in legal work but also for investment advice, according to Sven Kohlmeier, co-head of swissICT's legal commission and an IT specialist lawyer at Wicki Partners. He said he has been told by friends that some users rely on AI for investment advice and receive better stock suggestions than from their advisers.
The next step would be AI agents that execute trades independently, buying and selling shares. Confidence in AI-generated stock recommendations is already qualified by standard disclaimers stating that AI can make errors and that users should verify answers. Nevertheless, users place considerable trust in such systems because they appear unbiased and have no commission interest, a development that poses a challenge for human advisers.
Financial intermediaries that use AI applications should review the terms and conditions of model providers such as ChatGPT or Anthropic. Those terms effectively exclude liability and limit warranties, often in capital letters. Total liability is capped at the higher of the fees paid by the user in the previous 12 months or $100. The terms also set jurisdiction at the provider's location, in the United States.
As a result, legal enforcement against an AI provider is largely unavailable unless the intermediary has significant market power, financial resources or individual contractual arrangements. A claim against the provider is therefore generally difficult to enforce.
Businesses that use AI systems in their professional work remain liable for the outcome and for any defective performance. The explanation that an AI system produced the result does not, at present, relieve them of responsibility.
A Swiss Federal Court decision, BGer 4A_305/2021, which Kohlmeier considers transferable to AI systems, held that using tools such as a computer to automate market making does not constitute impermissible substitution under mandate law. Financial product providers may therefore use AI systems to fulfill orders, possibly even as part of their duty of care, but they remain responsible for the result.
Employees face additional obligations, particularly to protect business and trade secrets. AI systems are sometimes used without employer authorization, a practice described as shadow AI that creates problems in many companies.
To manage the legal risk, companies should adopt an AI governance strategy and internal rules governing AI use. Customer contracts should be reviewed to determine whether liability terms create an excessive gap when AI systems are used and whether transparency notices are required, including for copyright reasons. Model contracts, such as those offered by swissICT, can be used.
swissICT, described as Switzerland's largest ICT association, connects the IT industry, user companies and professionals. Its legal commission regularly publishes columns on technology-related legal issues.
