LegalTech: solutions to old and new challenges in internal investigations

August 2020 | EXPERT BRIEFING | LITIGATION & DISPUTE RESOLUTION

financierworldwide.com

Artificial intelligence (AI) and other machine capabilities closely associated with AI, such as machine learning (ML) and predictive coding, are not merely buzzwords in legal compliance advising. They are part of numerous LegalTech tools – the use of which has become standard in the e-discovery process of internal investigations – for example in an early case assessment, during the review or in the subsequent systematic analysis of the facts. Describing these developments as disruptive would be misguided and at least several years too late.

Today, it is not the technology that is potentially disruptive but external factors. With the regulatory framework for internal investigations subject to continuous change, this can result in an increasing number of requirements for how companies must structure their fact-finding processes, which thereby also determines the use of innovative technologies.

In addition, enforcement authorities are known to impose requirements on companies, either explicitly or implicitly, as the underlying expectations of cooperation. A closer look shows that LegalTech solutions may be a means for dealing with both old and new challenges arising in internal investigations.

Speed – fast and structured processing of large data amounts

LegalTech tools may help accelerate internal investigations and increase the overall efficiency of the process without resulting in a loss of precision. Many steps that formerly had to be performed manually can now be carried out by targeted automated solutions, thereby saving time and resources while, at the same time, providing greater flexibility.

The need for a speedier process initially results from the companies’ intrinsic motivation. By conducting their own, fast fact-finding processes and disclosing facts, enforcement authorities have the potential to complete investigations that would normally take years in a relatively short period of time and in a cooperative atmosphere. This also serves the company’s goal of getting out of the line of fire, avoiding reputational damage from coercive measures and negative press coverage.

In many cases, fast action is also needed in view of legal framework conditions. It is generally known from competition law that a company should inter alia present any incriminating evidence of the existence of a cartel ahead of the other cartel members in order to be eligible for immunity or a reduced fine. In practice, this leniency regime may result in a race between the cartel members and high time pressure in the fact-finding process.

The proposed German Act on the Sanctioning of Entities provides for monetary fines up to 10 percent of annual turnover for even the smallest criminal offences committed by employees or third parties in a corporate context. This is anticipated to become binding law in the near future. For companies doing business in Germany, this law would also create incentives to speedily complete fact-finding. The current draft states that companies must decide “in a short space of time” whether or not, and to what extent, they intend to cooperate once enforcement authorities have approached them in the context of a criminal investigation.

It must be noted that this requirement only applies to a particular form of mitigating sanctions, and is therefore not mandatory for a favourable outcome. In any event, to meet this requirement, companies will need to get the big picture from huge amounts of unstructured data as soon as possible. Going forward, conducting an early case assessment to ‘slice and dice’ data, and identify relevant events and persons, communication data and communication threads based on algorithms, can therefore be helpful, especially in complex cases.

The following aspect also shows that early case assessment can make all the difference. One of the core elements of cooperating with enforcement authorities in all jurisdictions is to actively support and assist public authorities’ investigations by voluntarily disclosing facts. While, for example, the US justice manual refers to the early disclosure of “relevant facts” among “Filip Factors”, the future German Act on the Sanctioning of Entities applies a more prescriptive approach. Similar to competition law, to qualify for the particular form of mitigating sanctions referenced above, a company must contribute ‘significantly’ to clarifying the matter being pursued by the enforcement authorities.

If this requirement is interpreted in the narrowest sense, the company’s aims may be jeopardised if enforcement authorities produce usable results quicker than the internal investigation. This would, in turn, depend on certain external factors, such as enforcement authorities’ eagerness to investigate or any tips provided by whistleblowers. But in this respect as well, it is up to the company to mitigate risks by using LegalTech tools in its internal investigation. With these solutions, even huge amounts of data that appear to be unmanageable from a human standpoint can be processed and analysed in a speedy and structured way, which might provide the company with the head start needed in this context.

Controllability – do not boil the ocean

LegalTech tools may also help control the scope of internal investigations. A tried and tested approach in this regard is to first define the subject of the fact-finding process and the concrete methods to be used within the framework of binding ‘terms of reference’. Another promising way is to use LegalTech tools that develop a clear focus at an early stage and limit the volume of the potentially relevant data, for example by clustering under the application of unsupervised ML, or by conceptual categorisation using supervised ML.

In a LegalTech context, controllability also has a self-referential aspect. Besides complying with the principles of procedural fairness, another general goal of internal investigations is to have them conducted in compliance with all applicable laws. This is an important aspect under labour law, trade secrets law, healthcare law (depending on the business sector concerned) and data protection law – for example where private data is handled in the course of e-discovery.

Regularly, minor breaches of law do not counter the mitigating effect of an internal investigation on the whole, at any rate if the misconduct was not deliberate. In practice, however, legal framework conditions influence the choice and use of suitable innovative technological tools to a significant degree, at least from a risk standpoint. To prevent advantages, such as the benefit of much faster action, from being undermined, the concept underlying the chosen LegalTech solution should ideally already be customised to suit the specific exigencies of individual jurisdictions.

Transparency – whitening the black box

Legal framework conditions also place certain demands on the transparency of LegalTech tools, especially their functioning and underlying algorithms. Put simply, the bonus for voluntarily disclosing facts discovered in internal investigations is granted because, by facilitating the fact-finding process, the company helps eliminate structural deficits on the part of enforcement authorities. White-collar crimes are one particular area in which authorities are often limited in their ability to detect potential violations of law, let alone investigate such violations.

To realise this legal policy approach, the disclosed findings from an internal investigation must provide added value to the authorities’ investigation. Enforcement authorities usually have wide leeway to assess this assistance. This will surely be impacted by the fact that most German authorities are still sceptical of the use of AI in e-discovery fact-finding processes. This may be due, in part, to the fact that there are no general rules or standards governing their use, as in many other jurisdictions. Moreover, there are no precedents or case law (either criminal or civil law) – as opposed to, for example, the UK, where Pyrrho Investments Limited v MWB Property Limited has advanced the matter. The decisive factor for German authorities’ reticence, however, is that many solutions relying on AI are essentially ‘black boxes’.

Black boxes are hard for the German criminal justice authorities to accept. At the core of this reticence is the aspiration not to leave the task of ascertaining the truth – which is the aim of any criminal proceedings – to an unknown and possibly unlimited algorithm. Some time ago, a high-profile government working group concluded that constitutional law demands that, when algorithms are used in criminal proceedings, their functioning must be fully transparent. If, however, judicial decisions are the non-verifiable result of a system based on artificial neural networks, the working group believes that this constitutes a lack of judicial independence and impacts procedural fairness. This assessment concerns the use of AI in judges’ decision-making process. However, based on reasoning rooted in constitutional law, concerns can be easily transferred to the preparation of a judge’s decisions and, therefore, also to the work of prosecuting authorities, which is also subject to the principle of procedural fairness. What is more, prosecuting authorities must investigate both for and against the suspect. For logical reasons alone, this means that the rationale behind LegalTech solutions used to obtain evidence must be known, including their strengths, weaknesses and limitations.

Accountability and requirements for lawyers

Ultimately, the requirements imposed on internal investigations should mirror, as closely as possible, those that prosecuting authorities impose on their own work. In turn, LegalTech solution providers must ensure that their technical models function and the underlying assumptions can be viewed and understood by third parties. The keyword here is ‘accountability’. Lawyers must be able to know the strengths, weaknesses and limitations of the LegalTech tools they use. Challenges are posed when solution providers and developers do not even know how their algorithms work. Deep learning programmes exist where this is already the case.

As a result, the services offered by law firms will increasingly require an understanding of, and interest in, LegalTech tools. However, this does not require a new type of lawyer, but emphasises the lawyer’s original task of fact-finding, with the difference being that, besides the facts, the focus is increasingly on the methods used to handle them. Another skill that will be of more importance in the future is a lawyer’s aptitude for acting as a moderator and mediator.

Lawyers must endeavour to dispel any scepticism on the part of authorities and to win their trust in the reliability of facts discovered in internal investigations. One of the key factors in this context is the ability to knowledgeably explain technical background. But lawyers must also be capable of making authorities understand that, going forward, internal investigations will only rarely be possible without the intelligent and customised use of AI.

Sven H. Schneider is a partner and Mathias Priewer is an associate at Hengeler Mueller. Mr Schneider can be contacted on +49 (0)69 17 095 143 or by email: sven.schneider@hengeler.com. Mr Priewer can be contacted on +49 (0)30 203 74 248 or by email: mathias.priewer@hengeler.com.

Sven H. Schneider and Mathias Priewer

Hengeler Mueller