In this insight, we consider the impact and use of Technology Assisted Review Systems (TARS), more commonly known as Artificial Intelligence (AI), or colloquially referred to as the "robot" to review data for complex investigations and disclosure purposes during criminal prosecutions.

Since the recent collapse of three high-profile criminal trials by the Serious Fraud Office (SFO), the debate surrounding disclosure failures has continued to escalate. On the one hand it could be argued that the failures are unique to those cases but, on the other, it could be said that it is the challenging disclosure regime in the English courts that is over-burdened with the ever-increasing volume of digital data.

SFO trial defeats and disclosure failings

In 2022, the SFO was subjected to two lengthy reviews commissioned by (1) Sir David Calvert-Smith KC into the Unaoil case; and (2) Brian Altman KC into the Serco trial. According to the Attorney General’s statement, the reviews were necessary to understand the failings of the SFO and to provide recommendations for the future handling of large-scale complex fraud and bribery investigations and prosecutions.

The lengthy reports provide the details of 2 trials, one in the oil and gas sector, that alleged several counts of bribery surrounding projects in Iraq, and the other, fraud by false representation regarding the contracts for the provision of electronic monitoring services for the government in the UK. Between them, the reports imposed 29 recommendations on the SFO. The core themes around those recommendations are on data handling during large-scale investigations, resource management and investment.

The recommendations are fair, but on their own they do not go far enough.

Investment in the public sector is key to enable it to deliver justice. On 24 April 2023, the Prime Minister and the Technology Secretary announced a £100 million budget for an expert UK taskforce to help build and adopt next generation AI (on top of £900 million investment into computer technology). This commitment at least demonstrates some interest in AI by the government, but this investment is not focused on the criminal justice system, where crime rates continue to rise.

High-level crime: fraud and money laundering

At the year end 2022, fraud was estimated to constitute 41% of all crime, with one in 15 adults and one in five businesses a victim of fraud (2017-2020).

The Economic Crime Plan 2023 presented by then Home Secretary, the Rt Hon Suella Braverman KCMP, and Chancellor of Exchequer, the Rt Hon Jeremey Hunt MP, highlighted the importance of investment to detect, prevent and enforce against fraud and money laundering. Government has ringfenced £100 million over four years with almost 500 new financial crime investigators dedicated to tackling money laundering and asset recovery. The delivery of this plan was supplemented by the Fraud Strategy 2023 that will target money laundering and cyber-related fraud offences. Action Fraud (only created in 2021) will be replaced with a new National Fraud Squad.

Aligned with the government’s objective to deter crime, is the new corporate “failure to prevent fraud” offence under the Economic Crime and Corporate Transparency Bill 2023, due to come into force later this year. It will be a defence for a corporate to show that it has adequate procedures in place to guard against fraud being conducted by an employee or an associate of the company.

Why are investigations so hard?

The sheer scale of data generated by individuals through smart phones and devices at offices, homes and vehicles, presents real challenges for any investigator.

Any basic data collection for an investigation goes far beyond emails to include social media posts and interactions, a variety of applications and messaging platforms such as Slack, WhatsApp etc and electronic devices. In a criminal investigation, interrogation of the Internet of Things (IOT) devices such as Ring doorbells, smart speakers and in-car cameras may also be required. All this data must be collected in a way that preserves the chain of evidence and processed for review.

The challenge is how to manage and effectively interrogate these enormous data volumes while maintaining the integrity of the source. At Dentons, we use one database to ensure that all key stakeholders such as general counsel for the corporate, counsel for individual witnesses, and cross-border lawyers all have access to one portal and one database, which is set to requisite access levels for the specific usage.

Prosecutor obligations

For any prosecution, the government agency must properly discharge its legal obligations to schedule and disclose all relevant material (applying CPIA 1996, the supporting Codes and the Attorney General Guidelines 2022). This exercise takes years to complete for complex criminal investigations which often allege historic criminal conduct committed over several years, if not decades, between various subsidiary companies across the world.

In His Majesty’s Crown Prosecution Inspectorate Report released in May 2023, the SFO has explained how it must prioritise cases to manage its limited resources, with one case being held in a queue despite it being ready to go to court. This means corporate and senior executive suspects are waiting years for their cases to be brought to justice before a court of law in England and recent events have shown that failures in disclosure can fundamentally impact the success of a prosecution.

AI and review platforms can be used to enhance the quality of the review performed during investigations and will drive efficiency of these lengthy cases.

Private sector data review

In the private sector, review platforms provide tools that expedite review and transform an investigation by targeting key material, while ensuring that all relevant material is clearly and accurately indexed and reviewed. In fraud investigations, it is crucial to make sure large volumes of data are managed and structured for processing and review.

Prioritisation software enables swift review of likely responsive data, enhanced by “clustering” and “email threading” which allow rapid navigation of large data volumes. Data can be illustrated through graphic visualisation (for example, communications between people and locations).

Deep dives are programmed for specific owner/custodians’ data (for example, email inboxes for targeted recipients), and some software can even provide sentiment analysis, highlighting stress points within communications.

Key take aways

  • Government agencies will need to consider how to safely deploy TARS to speed up the investigation and the disclosure process throughout the lifetime of a case.
  • Knowing which tool to use for a given task is essential when facing a complex investigation.
  • The key to any TARS is the clear definition of the final work product and the expectations of what these concepts can do and what their limitations are.
  • Dip-sample review and testing the TARS throughout is essential to maintain reliability.
  • The government has invested in the detection and prosecution of fraud, cyber and money laundering offences.
  • A new failure to prevent fraud offence will be in force by the end of the year. Placing more onus and financial burden on corporates to prevent crime by having adequate procedures in place.