In today’s ‘big data’ environment, the greatest obstacle to an efficient and streamlined merger clearance proceeding is the sheer volume of electronic documents that modern businesses generate. Fortunately, sophisticated AI and data analytics tools are available to extract insights and satisfy disclosure obligations in a timely and affordable manner.

The power and potential of such tools were discussed during TransPerfect Legal Solutions’ (TLS) 2023 event, The Future of EU & UK Competition Regulation. At the event, a panel of experienced lawyers and data analytics specialists shared their best practices, new developments, and the various approaches favoured by regulators in merger review proceedings.

This blog post summarises the two key takeaways discussed by the expert panel, which was comprised of Ekaba Davies, Managing Director and Head of eDiscovery, Operations & Analytics at Standard Chartered; Melina Efstathiou, Head of Litigation Technology at Eversheds Sutherland; Dan Meyers, President of Consulting at TLS; and Rajuan Pasha, Director of eDiscovery at TLS.

1. TAR Is Becoming More Common in UK and EU Merger Review Proceedings

Much of the panel discussion centred on the use of Technology-Assisted Review (TAR), an eDiscovery tool that uses AI to identify relevant documents after rounds of training by human subject matter experts. TAR is an incredibly valuable tool in merger review proceedings because it accelerates the eDiscovery process. Often, litigation document review requires the examination of hundreds, thousands, or even millions of files due to the broad scope of issues that regulators focus on when assessing a proposed merger’s likely impact on competition.

The panel agreed that the use of TAR is most common in merger review proceedings before US regulators, such as the Department of Justice and the Federal Trade Commission. Additionally, there was consensus that European regulators – specifically, the Competition & Markets Authority (CMA) in the UK and the European Commission (EC) in the EU – are increasingly more open and appreciative of the value of a TAR-driven eDiscovery managed review.

Because TAR is less common in Europe than it is in the US, Ms. Efstathiou reminded attendees that there is a need for greater transparency and negotiation with regulators in order to gain their buy-in and approval. As a result, reciprocal conversations should be had with regulators about how to define and implement TAR. Although the process of engaging the CMA and/or EC requires some effort on the front end of the review process, the ultimate savings in time and money are significant, as TAR regularly reduces the volume of documents requiring manual, human review by over 40%.

2. Trepidation Remains Around All AI, Including TAR

While the use of TAR is on the rise in Europe, each panellist acknowledged that there is still a significant amount of trepidation regarding the reliability and accuracy of AI programs. This hesitation is a result of ChatGPT and the well-publicised ‘hallucinations’ that can be produced by that specific category of AI engine. Mrs. Davies noted that many companies have AI Advisory Councils, which have to approve the use of AI in their organisations to ensure that AI is developed and used responsibly and ethically.

In response to AI-induced trepidation, Ms. Efstathiou explained that when it comes to TAR in particular, it is not a ‘black and white’ solution that definitively decides which documents are relevant and which are not. Instead, she stressed that the word ‘assisted’ is not just in the name but a core concept in deploying TAR. Therefore, it is meant to assist, not replace, human review.

Likewise, Mr. Pasha clarified that with TAR, even if there is initial prejudice in the results based on the dataset that was first used to train the technology, ‘eventually that prejudice is negated because the more you train the system, the more randomised it becomes in finding documents for you to review’. Finally, Mr. Meyers noted that remaining doubts over the accuracy of TAR need to be viewed in context, not in a vacuum. This means that the alternative approach to document review in the ‘big data’ era (i.e., a brute-force approach to linearly review all documents with a large team of humans) is even less accurate because humans lose focus over time, and a team of humans will have differing understandings and interpretations of the legal and factual issues, resulting in inconsistent decision-making.