Insolvency investigations: the possible effects of artificial intelligence on directors’ liability
As technologies investigating the causes of insolvency become more advanced, the question arises what purposes do new generation insolvency investigations have in the legal system and in particular in the context of directors’ liability.
Could new generation insolvency investigations contribute as a learning tool in preventing insolvencies? An important ambition of the proposed EU Restructuring Directive (COM(2016)723 final; ‘Restructuring Directive’) is for all Member States to have effective preventive restructuring and second chance frameworks. According to the EU Commission, a well-functioning insolvency framework should encompass measures to facilitate viable businesses to restructure and to continue operating. On the other hand, insolvency rules should channel enterprises with no chance of survival towards swift liquidation and provide a second chance to bona fide entrepreneurs.
One may argue that the effectiveness of the above mentioned frameworks is dependent on two key actors and the interplay between these actors. The first is the director who has a duty to restructure and to avoid grossly negligent conduct that threatens the viability of the business (art. 18 of the Restructuring Directive). The second is the practitioner in the field of restructuring and insolvency who has to ensure that his/her service are effective, impartial, independent and competent (art. 25 Restructuring Directive). These two actors meet intensely in the context of an investigation of the conduct and financial affairs of the debtor to establish the causes of the bankrupt’s failure (‘insolvency investigation’). I assume that insolvency investigations are an overlooked instrument that may be pivotal in supporting the effectiveness of the frameworks mentioned above. In some jurisdictions, such as the UK, it is well established that the trustee in bankruptcy has discretionary power to investigate the conduct and financial affairs of the bankrupt to identify the causes of the debtor’s failure (s. 289 Insolvency Act). The Bankruptcy Act in the Netherlands obliges the bankruptcy trustee to investigate the causes of a bankruptcy by reason of public interest (art. 68 Dutch Bankruptcy Act). Insolvency investigations have gained ground in Europe, in particular in complex insolvencies. More importantly, the use of new technologies have made insolvency investigations more sophisticated and cost-efficient. However, little is known about the potential of these advanced insolvency investigations as a learning device to prevent insolvencies as envisaged by the Restructuring Directive, and its potential effect on insolvency and directors’ liability regimes in general.
The expected aim of an insolvency investigation is to find the truth. In practice, the insolvency investigation is conducted by the trustee himself or, usually in larger bankruptcies, by a forensic accounting firm. In the area of forensic accounting, we can see a mutual agreement on a standard for investigating causes of insolvency. Forensic accounting experts recognise the importance of the distinction between the facts and the judgements about these facts. Moreover, forensic accountants objectively weigh the internal and external causes of the bankruptcy. They compare the benchmarks of the industry or market with the circumstances of the debtor. Their work is finished when they have the facts sorted out. The aim of this investigation is clearly to find the truth.
There are no formal standards for an insolvency investigation. Bankruptcy trustees agree on one standard only and that is that the aim of the insolvency investigation is to investigate possible liability claims. Schimmelpenninck created a protocol for insolvency investigations; the aim of the investigation is clearly the assessment of liability. However, bankruptcy trustees are usually of the opinion that they strictly separate the fact-finding investigation from the investigation of possible liability claims. This isn’t entirely correct. There is no evidence of a separate fact-finding process. Bankruptcy reports usually combine the fact-finding investigation and the liability investigation.
The question arises whether the bankruptcy trustee should be involved in an investigation of the causes of insolvency or whether the investigation should be based on independent research. Can the bankruptcy trustee independently and without biases investigate the causes of the bankruptcy when his ultimate goal is to create as much value for the creditors as possible? On the other hand, how do insolvency practitioners know what decisions they should be making in an insolvency investigation?
Big data and artificial intelligence
With the progress in information technology (e.g. cloud computing, electronic social media, analytics), businesses are moving into the era of Big Data. Along with this development, bankruptcy trustees and forensic accountants face huge amounts of structured data (e.g. general ledger or transaction data), unstructured data (e.g. e-mail, voice, free-text fields in a database) and non-traditional data (e.g. third-party watch lists, news media, social media) in an insolvency investigation. Artificial intelligence and big data analysis are increasingly being used to investigate complex and large insolvencies, for example in the bankruptcy of one of the largest retail companies V&D in the Netherlands last year. While data analytics tools are becoming more advanced and forensic accountants continue to professionalise and train themselves in using these (new) tools, establishing rigorous methods of culling and analysing the amounts of (un)structured data and transforming it into relevant information remains an ultimate challenge.
So how are artificial intelligence and big data analysis used in an insolvency investigation? Artificial intelligence is not clearly defined, decades ago we called the technology that automated most processes in an airplane artificial intelligence. Artificial intelligence is a computer science that makes a computer mimic human behaviour. AI still depends on us, people, and is programmed manually to derive information from a specific place and generate output. Every algorithm has an input and an output, the data goes into the computer, the algorithm does what the humans programmed it with and out comes the result. It is not human versus the machine but it is human plus machine versus the problem. If the code is programmed correctly, AI is completely transparent and the use of AI in insolvency investigations alone will not affect the position of a director. But how do we know the code is programmed correctly?
Concepts in law are open-textured, additionally with more open than closed strings or sets, to use mathematical language. Legal concepts, therefore, cannot be modelled by universally quantified conditions. We cannot put a number on a circumstance or on behaviour. Furthermore, law and legal concepts evolve. This makes legal concepts blurry, such as “serious reproach”, but this is what enables law to adapt to societal change. Unlike mathematics, as a result of the open texture, legal questions have competing, reasonable answers and they too evolve. Legal argument can be viewed as an exercise in competitive theory formation: each party forms its own conclusion using cases and other information that support its desired conclusions while at the same time minimising the opposing conclusions. Ideally, the truth will come out. However, if only one side, in this case the bankruptcy trustee, can use AI in creating these arguments, what does this mean for the position of the other party, the director? Is it fair that the trustee can omit information from its own investigation to come to their desired outcome?
Questions regard the objectivity, validity, and reliability of the research results may arise, in particular when the purpose of an insolvency investigation was not transparent, clear or concise beforehand. The potential of an insolvency investigation as a learning tool could then be grossly undermined. To be clear, the UK Insolvency Act and the Dutch Bankruptcy Act do not prescribe insolvency investigations from the perspective of learning, but more so from the perspective of combatting fraud and deterrence.
From the point of view of law making, it is exciting and problematic that a new device has been adopted by which trustees have gained additional power to scrutinise directors’ behaviour in the pre-insolvency stage. One could argue that due to this new generation investigations, insolvency practitioners have a huge lead of information, in hindsight, which can affect the balance in the burden of proof in a potential directors’ liability proceeding. Do bankruptcy trustees utilise the new generation insolvency investigations as a learning device or as a liability-discovery device?