Impact of Artificial Intelligence on corporate management

Hardly any other subject area in law is currently as pervasively discussed as the “Digital Revolution”. Terms such as “Artificial Intelligence (AI)”, “Big Data” or “Internet of Things (IoT)” dominate the headlines in the media, both in the tech and finance eco-system. In particular the impact of AI on the management of corporations is a newly emerging application of digitization, which bears considerable legal challenges for companies.

Increasing importance of digitization for companies

As early as 2014, Deep Knowledge Ventures, a venture capital company based in Hong Kong, appointed an algorithm called Vital (abbreviation for “Validating Investment Tool for Advancing Life Sciences”) to its board. Although it fulfills a purely observational function, it was the beginning of the introduction of algorithms into corporate management. According to the current state of technology, it seems possible that in the near future AI could make management decisions or (partially) replace management. This article will discuss, whether companies are legally permitted or even bound to use AI.

Use of AI in corporate management

The integration of AI into corporate management poses new challenges to corporate law. These result primarily from the fact that the framework for management decisions has so far been tailored exclusively to human decision-makers and their responsibility. The central question is therefore whether and how AI phenomena can be integrated into this system. In order to answer this question, one has to distinguish between the use of (purely) information-providing AI, which an increasing number of companies has already introduced to their management, and the use of decisive/executive AI, which is still of a rather visionary nature.

Information-providing AI

By information-providing AI one understands the use of partially automated systems for the acquisition and evaluation of information as well as for the acquisition of knowledge and decision recommendations. According to German law, the following questions arise during their use in corporate management:

• The first question is to what extent the board of a stock corporation is already obliged to use such systems pursuant to sec. 93 para. 1 sent. 2 of the German Stock Corporation Act. This problem arises equally for the managing director of a limited liability company as well as for the supervisory board, to whom sec. 93 para. 1 sent. 2 of the German Stock Corporation Act applies as well. According to the therein regulated Business Judgement Rule, the board must act on the basis of appropriate information for the benefit of the company. In times of Big Data, the management will have to take into consideration that more comprehensive available data will increase the requirements for the assumption of a sufficient factual basis. Although a fundamental obligation to include an information-providing AI in management decisions is currently predominantly rejected in German legal literature, it is advisable to use AI in cases where a diligent decision can only be made with the help of AI. Therefore, it is important to keep a close eye on future legal and technical developments.
• When using information-providing AI, the management – similar to a delegation to source information to employees – has a monitoring obligation. It must evaluate and monitor the method the AI system works. To this end, the management must have at least basic knowledge of the way the AI system functions to prevent that it becomes a “black box”. In addition, the information of the AI must be subjected to a plausibility check, for which the “ISION” principles of the Federal Court of Justice (BGH) for the advice of experts (BGH NZG 2007, 545) apply according to German legal literature.
• When assessing and including AI in the decision-making process, the management must take into account the fact that the information determined by AI, if included in a decision, is legally attributable to management. The information advantage resulting from the use of AI therefore also entails certain disadvantages. Pursuant to the knowledge organizational obligation (Wissensorganisationspflicht) established by the Federal Court of Justice (BGH) (BGHZ 132, 30), a company is attributed such knowledge that is typically internally recorded in its filing and could be retrieved in the concrete situation with reasonable effort. When AI is used, this reasonableness threshold may tend asymptotically towards zero due to the amount of data available.

Decisive/managing AI

Decisive AI refers to the use of fully automated or even autonomous systems that make decisions independently. As far as the execution of “simple” management tasks is concerned, there are no objections to the use of such systems. Nevertheless, the management has to fulfill the already mentioned monitoring obligations. When delegating fundamental management tasks, a distinction must be made between fully automated and autonomous systems. The final decision must always remain with the management. The management is fully responsible regardless of the use of AI. Within this framework, the use of fully automated systems, the criteria and limits of which are defined by the management, is already possible today. The use of autonomous, independent systems, on the other hand, is impermissible. Errors or any fault on the part of the AI cannot be attributed to the management due to a lack of legal personality. However, the management itself can be held liable for the general use of such systems or the lack of its control or monitoring. Such liability risks can be prevented by careful handling of decisive AI. The German legal literature tends to apply the provisions of sec. 80 para. 2 of the German Securities Trading Act, which refers to the use of algorithms in securities trading, but which, in the absence of other regulations for the use of AI, could be considered as a general legal concept.

The use of managing AI, i.e. as an organ of a stock corporation or a limited liability company, is ruled out de lege lata for lack of legal personality. It is opposed by the clear wording of sec. 76 para. 3 sent. 1 and sec. 100 para. 1 sent. 1 of the German Stock Corporation Act as well as of sec. 6 para. 2 sent. 1 of the German Law on Limited Liability Companies.

Conclusion

Although the use of AI in corporate management is already technically possible in different ways, the final responsibility still lies with the board or the management. If AI is used to minimize liability risks or in case of future legal obligations for the use of AI, corporations must always observe the aforementioned monitoring obligations. Otherwise, the introduction of technical progress threatens to become a liability risk.