The Serious Fraud Office (SFO) has, unusually, allowed BBC News recent access to their new weapon against crime. They are trialing Artificial Intelligence technology that can analyse a mass of electronic documents significantly quicker than human intelligence officers.

Many businesses are increasing their productivity by using AI to automate systems and save on staffing costs. If your business is researching how to integrate AI into your processes, you are probably entitled to R&D tax relief. Everything involving AI, from off-the-shelf chatbots to cumulative machine learning, is being tried and tested across many different industries. The tech itself is truly innovative, so the chances of your project being eligible for R&D tax credits is very high.

What does the SFO’s new technology actually do?

The OpenText Axcelerate system creates ‘hypergraphs’ which identify connections within thousands of emails. The Enron Corporate Fraud Case is a training context in which SFO officers can learn how to use this cutting edge technology. It hones in on evidence which links two individuals involved in fraudulent behaviour. It analyses emails senders, recipients, everyone copied in and those without anyone copied in. The AI technology also makes links between content and specific language used by individual senders. So investigators can pinpoint words or phrases that are of importance.

As Edgar Pacevicius, an SFO investigator, explained to the BBC: “We normally see a lot of euphemisms – there’s a lot of potential deception about the way people do corrupt activity. What we’re trying to achieve is to find an intelligent technological solution that will allow us to not only identify those phrases but everyone involved.”

How often is this new AI tech used in SFO investigations?

The new system has only been used since March this year in three active investigations. It is still very much in the trial phase, although the intention is for it to be available to use in all corruption, bribery and fraud cases across Northern Ireland, Wales and England.

AI assisting with identifying information protected by legal privilege

While investigating a case, there is some information that the SFO are not allowed to view because it is protected by ‘legal privilege’. There is another new AI system called RAVN that is being trialled in order to speed up finding such information. A computer running RAVN can sort through 600,000 documents a day, identifying those subject to legal privilege as it analyses them. A barrister can get through 300 per day. The speed of analysing any evidence has a direct impact on how long an investigation takes to complete and bring suspects to trial.

The RAVN system was used by the SFO to support their four-year long fraud investigation into Rolls Royce. They had 30 million individual documents to inspect, filtering out those protected by legal privilege. Imagine how many more years it may have taken without the use of RAVN.


AI is also being used in conjunction with CCTV footage, again, to speed up investigations. SeeQuestor the tech company, has invented a system which processes footage from multiple cameras. It matches faces of suspects, victims and witnesses and can help map out there movements as they appear on different cameras.

The technology helped during the murder investigation of 12 year old Tiahleigh Palmer in Australia. Their police force was faced with 21,000 hours of footage to work through, trying to find out what happened to her. By using SeeQuestor technology, the two detectives needed only four days to process all of the data.

It has been used in Britain on six cases, including the high profile disappearance of Airman Corrie McKeague.

If you work in the field of CCTV development, this isn’t the only kind of innovation that would be considered for R&D tax relief. For example, increasing the remote functionality and accessibility of CCTV systems from different mobile devices. The pace of development in this field is fast and any work you do to assimilate new technology is likely to fall under the definition of R&D for tax relief purposes.

Why isn’t AI technology being used more in law enforcement?

Head of the SFO’s Bribery and Corruption Unit, Camilla de Silva, told the BBC: “I’m a lawyer – lawyers are inherently conservative sort of people. We are going to test and retest and make sure that it’s an appropriate use of technology before we let it out anywhere near a criminal court.”

Anything that impacts on the collection of evidence must be one hundred percent reliable in order to be truly valuable during court proceedings. So AI technology is being cautiously considered by those trialling its potential.

There are many other concerns connected to the police use of Artificial Intelligence. Are our civil liberties fully protected? What is the margin for error within the AI systems? How do we best protect against miscarriages of justice brought about by the use of AI during investigations? That’s just three of the many questions that must be answered alongside our pursuit of cutting edge crime fighting weaponry.


Jamie Smith