Update on cryptoactives and artificial intelligence

(Translated by DeepL)

FINMA has just published its annual report, which reviews the year 2022 and outlines the priorities for supervision. These include issues and challenges posed by the current geopolitical context and inflation, money laundering supervision, climate risk and anti-money laundering supervision, and digitization of the financial sector. This last point is the subject of this commentary, which discusses FINMA’s practice in relation to cryptoassets and takes stock of artificial intelligence (AI) in the financial sector.

FINMA notes an expansion in the range of services offered by Swiss banks geared towards cryptoactives. In particular, it mentions staking services, assistance with the issuance of non-fungible tokens (JNF), or the tokenization of real assets. It also mentions a certain interest on the part of financial players for authorization as a trading system based on TRD, although no concrete application has yet been filed. In this respect, the report informs us of certain solutions and practices that have emerged from exchanges between FINMA and certain players. For example, one of the questions that arises is who would be responsible for the custody and oversight of TRD-based securities portfolios, if the custody of assets is based on a public blockchain. FINMA points out that, in such a case, the TRD-based trading system must, on the one hand, ensure that delivery against payment runs smoothly, and on the other hand, regularly check the operating capacity of the public blockchain and the level of participants’ deposits on the basis of public information. This would not exempt the system from certain obligations.

With regard to non-fungible tokens (JNF), FINMA reiterates its well-established practice of substance over form. The analysis focuses as much on the technical vehicle as on the asset or right represented in the token and its economic function. The will of the parties, the conditions of issuance and the terms of use of the trading platform may provide information on the rights of the owners of the JNFs. Furthermore, FINMA does not rule out a certain fungibility of JNFs.

Still on the subject of cryptoassets, FINMA mentions exchange trade products (ETPs), which are debt instruments that replicate the performance of an underlying asset. In recent times, ETPs have been launched with cryptocurrencies such as bitcoin as their underlying assets. FINMA points out that ETPs can be considered structured products within the meaning of the LSFin. As such, the corresponding provisions of the LSFin are applicable and must be complied with.

Digitization of the financial sector also involves greater use of artificial intelligence (AI). The report takes stock of the use of AI in the financial sector. At present, AI is mainly used in the front office and to optimize internal processes. The fields of compliance, risk management and system monitoring – for example, the monitoring of IT security perimeters – are not to be outdone, and are also using AI. We also learn that FINMA’s work on prudential expectations targets the areas of governance and accountability, transparency and explainability, equal treatment and the robustness and reliability of AI applications. It is interesting to note that these aspects are given particular attention in the European draft AI regulation and form the core of this proposal (cf. cdbf.ch/1181/).

Finally, FINMA points out that AI entails risks. By way of example, it cites the development of applications by AI but without human intervention, or unequal decisions taken by an AI. This last example is a crucial issue in the regulation to develop AI without bias. However, the report fails to mention other potential risks for financial markets and reporting institutions. Indeed, in recent months, so-called generative AIs – such as ChatGPT or Midjourney – have enjoyed dazzling success with the public. While their use can bring considerable benefits for everyone, they also entail considerable risks. In addition to data protection, we believe that generative AI poses a more fundamental challenge to our society and to the financial sector, namely public misinformation and market manipulation. Thanks to this kind of AI, fake news about companies or leading figures can be created in a few clicks and spread on social networks. Given the sometimes erratic reactions of financial markets to rumors spread on social networks, the combination of AI and social networks can have harmful effects on listed companies and financial markets (see thesis currently being written on the subject). FINMA should – in our view – address this issue in the near future.