The Cabinet Office has announced plans for a number of AI discovery projects over the next 12 months to identify new ways to detect fraud against the public sector.
Minister of State at the Cabinet Office Baroness Neville-Rolfe said the first project will use AI to identify entities registering and bankrupting successive companies to avoid paying debts, known as ‘phoenixing’.
All of the projects will tested by the Public Sector Fraud Authority (PSFA) in accordance with the Government’s Generative AI Framework.
The move follows the provision in the Budget of a £34 million investment to build the next generation of counter-fraud technology.
Platform upgrade
It has been accompanied by the announcement that the Government’s AI powered fraud detection tool – the Single Network Analytics Platform (SNAP) – has been upgraded with thousands of new sanctions and debarment records.
Three new datasets have been added: 18,000 UK and US sanctioned entities, including those introduced following Russia’s invasion of Ukraine; 1,000 World Bank debarments, that are deemed ineligible for World Bank contracts; and 647,000 UK dormant companies that do not have any income.
The Cabinet Office said that the new data will provide more support to public sector organisations in detecting fraudulent claims on public funds through contracts, grants and loans.
SNAP was launched by the Public Sector Fraud Authority in 2023 as part of its £4 million partnership with the UK based AI tech leader Quantexa. The tool will continue to be regularly updated with new datasets.
Tech on front line
Neville-Rolfe said: “Criminals should be aware that we’re putting technology on the front line to detect fraud and protect taxpayers’ money.
“Adding sanctions and debarments data to our AI counter fraud tool will help us identify organised networks stealing from the public purse. And over the next year, we’ll be running new discovery projects to identify even more ways to use AI to detect fraud against the public sector.”
In January the PSFA revealed that £99.5 million of pandemic-era fraud had been identified through a data sharing pilot enabled using the Digital Economy Act 2017. This is one of more than 100 pilots it has run with 70 local authorities and 17 government departments since 2018.
