eDiscovery technology and associated processes, including the use of algorithms and artificial intelligence (“AI”), have become a vital tool helping teams to navigate ever-increasing data volumes, including the additional complexity of a greater variance of data sources. In the U.K. disputes space, the use of these technologies and approaches is governed by the U.K. Courts. Specifically for the Business and Property Courts, guidance can currently be found in the CPR Practice Direction 51U, referred to as the ‘Disclosure Pilot Scheme.’

The benefits of eDiscovery technology are clear to see in the litigation arena and the technology is now widely accepted and even encouraged by U.K. courts through the Disclosure Pilot Scheme. In particular, the use of algorithms to give greater insights into our data and the use of AI to assist with any review process are fundamental in many cases. For example, the technology can be used to:

• Group together documents similar in conceptual content by looking for patterns in the individual words used across them – an advancement on the traditional approach of keyword searching. The results of this type of analysis can be visualised in ‘clusters’ allowing a user to graphically interrogate their document universe.
• Understand highly correlated terms, synonyms or strongly related terms in the document set. Users can explore how different language is used to express the same or similar concepts allowing them to create a more effective keyword set. This in turn leads to fewer false positive documents for review.
• Continuously prioritise documents for review, based on coding decisions made in real time (Continuous Active Learning, “CAL”). A CAL model can predict which documents may be relevant to a case and complete the document review process. Using a CAL model to supplement human reviewers reduces the overall time, resources and costs of such an exercise. Statistical evidence to defend the CAL model approach is produced and has been widely accepted by U.K. courts.     
• Enhance compliance by monitoring anti-competitive language or behaviour. AI models can review the language and content in email systems or chat messaging systems in real time to detect any potential issues at source.

eDiscovery Technologies in Competition / Anti-Trust Cases and Litigation

In regards to cases and reviews conducted by competition authorities, the data challenges are often similar in nature to those faced in commercial disputes. Therefore, the technologies outlined above also play a key role in this context. By their nature, competition / anti-trust cases and follow-on litigation are often ‘high stakes’ and cross-border, involving the handling of large data volumes and with an emphasis on uncovering important evidence as quickly as possible.

Merger Reviews

Competition authorities review merger transactions in order to determine whether they are likely to result in a lessening of competition in the relevant markets, and bring about consequent harm to consumers – in the form of high prices, deterioration of quality or harm to innovation. In case of concerns about the likely effects, a transaction can be blocked or significant remedies may be imposed in order for it to be approved.

To inform their (forward-looking) assessment, competition / anti-trust authorities look at oftentimes complex economic analyses of the likely effects from the merger. This involves collecting masses of data on the merging parties, their competitors, and the market more generally. This data will need to be extracted from a variety of sources, including live or archived databases, studies, presentations and files. Over the years, this has become an increasingly involved and lengthy process.

Equally, companies involved in a merger review process will be expected to disclose internal documents to the competition authority. The authority will use these documents to better understand the market and the competitive conditions, as well as the rationale for the transaction. These may be communications between senior members of staff, board presentations, marketing material and business projections. They may be stored in a variety of systems such as email systems, laptop / desktops, document management systems, file servers, cloud-based applications and mobile phones.

Cartel investigations

Competition authorities investigate suspected agreements between competitors, or market participants more generally, that have the effect of dampening competition and hurting consumers. An investigation can be triggered by a whistleblower, a complaint, or ex officio. The focus of those investigations is on establishing the illegal behaviour, through communications between firms, agreements on pricing, market sharing arrangements or bids in tenders. Fines can be up to 10 percent  of the implicated group’s turnover and may be followed by damages claims.

The competition authority will send elaborate requests for information that cover an extensive and wide array of material, documents and data. Moreover, in many cases, at the start of its investigation, the authority will raid the premises of the suspect companies (and possibly the domestic premises of key business individuals) to collect evidence to inform its investigation.

In response, many companies start their own investigation, not only to show a willingness to co-operate with the authorities but also to more fully understand potential risks and their exposure so as to better mount a defense. Many companies will mirror the data collection process conducted by the competition authority and fast-track the processing, analysis and review of the information to anticipate the findings before the investigation itself is complete. Early detection allows companies to assess whether they should apply for leniency or engage in settlement discussions with the authorities.

Further, as the investigation progresses, additional evidence may also be collected as new data sources or additional relevant personnel are identified. In these circumstances, the ability to de-duplicate new data against existing data is key. Existing relevant documents already identified can also be used as a ‘seed set’ to predict which documents are the most relevant from the newly identified data sets.

Although often a by object infringement, there is also the need for competition / anti-trust authorities to document and measure the effect that alleged agreements between companies have on the market and on consumers. This leads to an additional data mining exercise whereby transactional, financial and market data are processed and analysed. Given that these investigations generally look back many years, the data may be – again – stored in several sources including off-site archives.

Abuse of dominance

Part of a competition authority’s remit is to also investigate unilateral anticompetitive conduct by dominant undertakings. This can take many forms including exclusionary or exploitative conduct. Authorities will look at information held or exchanged within the company to understand their financial (cost and pricing) data, their view of the market and rivals, and their commercial strategy. Much like in the cases above, there is therefore a need to take a detailed look at operational and transactional data, documents, communications, internal assessments and strategy papers where eDiscovery professionals and the processes detailed can ensure an efficient review process.

What is the timing of involvement?

Many of the examples discussed so far are reactive in nature – for example, responding to a request during a merger process or instigating an investigation following a dawn raid. However, there are always benefits to engaging in a discovery exercise early. This helps the companies concerned anticipate the authority’s or adverse party’s findings, and advance the appropriate argumentation.

Moreover, eDiscovery technology and the use of AI is also playing a larger role in the compliance function of many businesses. Systems using AI models can monitor communication and transactional systems in real time and flag any competition / anti-trust concerns early. This allows the company to conduct any internal investigations at the earliest point of detection, aiding their position if an authority did conduct their own investigation or if there is a need to consider a leniency application.

The compliance function can provide the AI model with feedback on a particular flag allowing it to continually improve its understanding of potential risks. These monitoring systems can be looking at millions of transactions and communications per day and they can reach accuracy levels of over 95 percent (in other words, 95 percent of the flags raised by the AI model being deemed important in the context of compliance monitoring).  

Summary

While there is no clear guidance from competition / anti-trust authorities on the use of eDiscovery technology, and in particular the AI methodologies, to assist with what was traditionally human review work, many of the larger authorities have endorsed those technologies. Given the need to process voluminous data and information, technology-assisted collection and analysis is both being accepted and encouraged, as long as the processes are clearly defined by the company and their eDiscovery partner. As with all technology approaches, transparency of process is key; as is the experience of any technology partner.

With AI continuing to have a greater effect on our work and personal lives, it is only a matter of time before the use of technologies and processes described above are whole-heartedly accepted. We would anticipate that before that milestone is reached, courts and regulators will continue to take a closer look at how these technologies are applied in the real world. As their adoption increases, courts and authorities will want to develop a deeper understanding of their usage, and answer questions of legality and fairness in how they are applied.