Cyber risks

Another issue that merits regulatory attention, not only in regard of robo advice, but rather concerning the fintech phenomenon as a whole, is the problem of cyber risk. The cyber risk susceptibility of financial market actors is high due to the level of interconnectedness. Given that fintech start-ups are frequently not in possession of a proper cyber-security system comparable to those of big financial institutions, cyber risk is an issue that merits further attention. On the other hand, in case of more diversity and less concentration in the market that may come along with the rise of fintech, a singular cyber-attack may be less systemically relevant where it can be contained to the attacked entity or context.

Artificial intelligence

One factor that plays a great role in the future development of robo advice, bearing great potential for its quality, but also posing serious risks, is the increasing use of state-of-the-art Artificial Intelligence (Al) and machine learning solutions. These technologies can be fed with stacks of data consisting of parameters about individual investors, such as their credit history, employment history, assets, purchasing history as well as data that stems from social media, for example, Facebook or Twitter. Also, the algorithm may use data about macroeconomic parameters, such as market movements and collective behaviour during volatility.

With datasets about the individual, robo advisors using Al could design a portfolio more tailored to the individual preferences, that is, give a more customised advice. Some commentators see the inclusion of Al and machine learning as the yet missing piece in the puzzle that will allow robo advisors to widely replace human financial advisors. Obviously, it could present an opportunity to overcome the problems associated

Robo advice - legal and regulatory challenges 205 with the use of a questionnaire that we described earlier. The robo advisor could use various datasets to verify clients’ answers to the questionnaire and thereby address certain biases prevalent in the answering process.[1] Even more audaciously, future robo advisors might refrain from using a questionnaire altogether and instead solely rely on data on the user that they retrieved from other sources. The interaction with the client would thus be limited to exploring risk tolerances and investment goals. Such an approach would potentially make the advice more precise, but certainly more convenient and faster. Also, Al could enable robo advisors to offer a more comprehensive product to the customer and automate parts of the financial advice service that are to date reserved for human financial advisors. Feeding the algorithm with macroeconomic data could enable it to even anticipate market movements and the occurrence of shocks and to better estimate certain risks. On the other hand, since an algorithm is always ‘trained’ on historical data, it might fail to anticipate utterly novel categories of risks. Relying on the algorithm irrespectively may pose an idiosyncratic risk: algorithms using machine learning develop their own dynamics, which can lead to the problem that commentators commonly refer to as ‘black boxes’ in decision making. This describes a decision made by algorithm which humans find difficult or impossible to trace and understand. Such black box decisions pose difficult questions in relation to liability, auditability and - of course - regulation. Also, as a consequence of unpredictable and unexpected decisions, in the absence of data (and a better explanation) market movements may be ascribed to Al and interpretation of market shocks may therefore be hampered.

On balance, robo advice combines several important benefits, for investors as well as for the financial market as a whole. However, many risks are still unresolved, notably the risk for investors to receive unsuitable advice and the market risks of increasing volatility and potential flash crashes. For both, risks and benefits, we may see some interesting developments with Al and machine learning in the near future. Again, more data and time are necessary to comprehensively assess the phenomenon of robo advice exhaustively.

  • [1] See also Wedlich (n 50) 227f. 2 As described earlier (see item 3.4. of this chapter), these parts largely consist of relationship management and coaching. 3 See also FSB (n 65) 34. 4 See for example Rosov (n 81). Also EBA (n 45) sees that risk specifically in respect to robo advice (see on p. 21). 5 FSB (n 65) 26. 6 FSB (n 65) 30. 7 See for example OECD (2009) ‘Policy Framework for Effective and Efficient Financial Regulation’ 17; Armour and others (n 15).
 
Source
< Prev   CONTENTS   Source   Next >