Provincial court hears the AI enigma
One of the characteristics of the judicial system is that the plaintiff's place of residence determines the place of jurisdiction for proceedings. The Lüneburg Regional Court now has the task of hearing a case at the intersection of crypto trading and artificial intelligence (AI).
The background: Some investors who have suffered losses as a result of the burst crypto bubble have gone to court. In Lüneburg, for example, a private investor is accusing the crypto investment platform Invao of fraud: It had not used artificial intelligence (AI) as advertised in prospectuses.
Verdict expected at the end of March
The founders of Invao, led by Frank Gessner as the defendant, reject the accusation. The platform's investment algorithm, which is programmed for various trading strategies, is indeed equipped with AI functions.
In order to clarify whether there is a claim for damages, the Regional Court of Lüneburg must deal with the question: when is an AI really an AI. At the hearing, which was reported on by the industry portal Finance Forward, witnesses testified that trading on the platform was indeed largely automated. The AI was fed with historical data, a method known as backtesting.
Dispute over training intervals
However, it remained unclear how often the AI was trained. Witnesses stated that this took place once a quarter. This seems like quite a long interval, as today's AI founders report that they train their investment algorithm once a week. Invao founder Gessner initially left an enquiry about the quality of the AI unanswered.
A lot can happen in a trading week: Market participants change their positioning, central banks tighten interest rates and thus influence the prices of almost all asset classes, and companies present business figures or other relevant information. A weekly training session with data from the past, therefore, seems entirely appropriate.
Self-learning AI
The plaintiffs take what they see as too long a training interval as an opportunity to deny that the Invao software is a self-learning AI. This would require the algorithm to be able to adapt its behaviour based on the knowledge gained. Self-learning systems use databases to recognise patterns and regularities. They use this to develop solutions after the programmer has determined how this should be done.
Experts differentiate between machine learning and AI. According to the Fraunhofer Institute, this intelligence can be based on programmed processes or generated through machine learning. In machine learning processes, an algorithm learns through repetition to fulfil a task independently by recognising the structure of the data on its own.
In layman's terms
So far, so complicated. But the IT laymen on the bench need to form an opinion: Did Invao really solicit investor money with false promises? To answer this question, the judge would actually have to obtain an expert opinion. The proceedings will continue on March 20 with case number 5U290/22, and an expert opinion could also be requested with an order for evidence, according to a spokeswoman. The parties to the dispute are expected to submit statements by the March date.
Like a high-frequency trader
Invao Chairman Frank Gessner also spoke to Börsen-Zeitung about the business model in 2019: After an initial stage of active portfolio management, day-to-day business is conducted via automated trading "using AI-supported systems".
According to the founder, the platform utilises "a wide range of trend signals and functionalities, as are being used by high-frequency traders". As auditor, KPMG ensures that the funds invested in the 3,000 to 5,000 trades per day are appropriately utilised.
FTX debacle was not foreseeable
In terms of compliance, Invao was, therefore, fundamentally well positioned. And it can be assumed that anyone who invested there was aware of the risks of the crypto market. But nobody could have realised that Invao would be shipwrecked by parking its assets on the now insolvent US exchange FTX.
However, investors who were interested in diversification from individual cryptocurrencies could also have invested in the CCi30 benchmark index. It tracks the performance of the 30 largest assets on the blockchain market.
Additional return opportunity
The appeal of an investment lies in the additional return opportunity offered by the Invao portfolio, with 60 possible target values compared to the benchmark. And those who opted for the Invao portfolio – with AI or without – consciously took an investment risk, as with any other investment decision, for which there were alternatives.
