Illuminating shadows: tackling fraud and error in government

Photo by Lennart Wittstock via Pexels

During this webinar, experts shone a spotlight on fraud and error in the public sector, focusing on how to measure it, how to combat it – and how AI is changing the game

Estimates of money lost to fraud and error through orchestrated attacks on government programmes, incorrect payments, tax evasion and similar are eyewatering – and by all accounts, the estimates are likely considerably lower than the true scale of the problem.

The COVID-19 pandemic revealed just how vulnerable governments are to targeting by state actors and organised crime groups, as well as to welfare fraud. Now, at a time when public spending in most countries is stretched, preventing fraud and error and recovering stolen or misplaced funds is an increasing priority.

In a recent Global Government Forum webinar, experts from the supreme audit institutions of the US and the UK, Nigeria’s chartered institute of fraud examiners and the webinar’s knowledge partner Unit4, talked about how government can understand fraud threats, measure the problem, develop and implement counter-fraud tactics, and enhance cross-border collaboration.

Getting a grip on the numbers

Panellists kicked off the session by discussing the extent of taxpayers’ money lost to fraud and error and the progress and challenges around transparency and reporting.  

Joshua Reddaway, director of fraud and propriety at the UK’s National Audit Office (NAO), highlighted that by latest estimate the UK government lost between £55bn and £81bn (US$72bn-$106bn) through fraud and error in 2023-24.

Of that, about £39bn (US$51bn) relates to tax fraud and avoidance, and £11bn (US$14bn) to benefits-related fraud and error.  

COVID was a “real wake-up call” in terms of alerting  government to the scale of the external threat and the realisation that fraud controls required more thought, he said. “It really drew out just how unprepared most departments were for it.”

The NAO estimates that about £10.5bn (US$14bn) was lost from the government’s response to COVID, mostly through the Bounce Back Loan Scheme which provided emergency business finance.

Since then, the NAO has been working to get a handle on the numbers, including through extended sampling and asking for more information about government transactions than has traditionally been required

Reddaway said that as well as reporting improvements – reporting rates have doubled in recent years – more robust risk assessments and control designs have also been implemented, but he acknowledged that it’s still a work in progress.

Read more: UK government flows sanctions data into AI-powered fraud detection tool

Hundreds of billions

Linda Miller is president of the Program Integrity Alliance, a nonprofit organisation that aims to enhance integrity and fraud prevention in US government programmes.

She said the amount of money lost by the US federal government through fraud and error “hovers” around US$220bn to US$250bn a year, but is thought to be underreported.

Government agencies are required to estimate the ‘improper payments’ – covering fraud and error – they’re making based on a “very specific formula” under the Improper Payments Reduction Act, but as she noted, the US Government Accountability Office (GAO) “has routinely found that agencies struggle to do that effectively”.

Indeed, GAO undertook what she described as a “really ambitious” piece of estimation work using the Monte Carlo simulation of probabilistic modelling and estimated that the US federal government loses between US$230bn and US$521bn a year through fraud alone.

This number is “somewhat controversial”, she said, with some observers believing it to be far higher than reality and others, like her, who think it is likely too low. The “short story”, she said, “is that the problem is enormous, and we’re really challenged in measuring it well”.

Miller described some of the state sponsored and organised crime group activity directed at the US, including the targeting of the Supplemental Nutrition Assistance Program by Romanian criminals, identity theft-based fraud in student loan programmes perpetrated by North Korea, and Chinese and North Korean targeting of its Unemployment Insurance programme.

During the pandemic, fraud committed against the Unemployment Insurance programme could have amounted to as much as half a billion dollars, Miller said.

She echoed Reddaway’s point about the pandemic waking government up to the scale of fraud and error and said that, although GAO’s most recent estimate has drawn more attention to the problem, “I would argue that we haven’t done enough as a government-wide effort”.

Read more: US pandemic accountability committee gets funding to continue work to 2034

Disconnected data

Bryce Wolf, director of strategic growth at Unit4, agreed that the scale of the problem is bigger than people realise and that most public sector organisations are “probably just scratching the surface of it”.

He said governments have the desire and intent to reduce fraud and error but there is one key and consistent challenge, even for digitally advanced economies and organisations: that of visibility.

“You can’t measure what you can’t see, and many [governments] are really still struggling with disconnected data across various departments,” he said.

He highlighted Unit4’s State of the Digital Nation report, in which around 65% of public sector leaders reported that the information they need could be easier to access.

“That lack of access makes it hard to get a true picture of fraud and error”, he said, adding that when governments invest in integrating systems and real-time insights, “that’s when they’re able to take that measurement and turn it into something meaningful and actionable”.

In Miller’s view, there are legal and institutional limitations around data which also need addressing. In the US, “the privacy protection laws that govern how data is used are really important to protect privacy but they’re also anachronistic. They were built before the digital environment we live in today and so we believe strongly that those privacy laws need to be revised”, she said.

Read more: Canadian government steps up anti-fraud efforts including new Financial Crimes Agency

Cross-border collaboration

Johnson Oluata, vice president of Nigeria’s Chartered Institute of Forensic & Certified Fraud Examiners, outlined the extent of the problem in his country and the global nature of fraud committed against governments and need for cross-border collaboration.

In Nigeria, he said the problem of fraud is exacerbated by the level of poverty in the country, by issues around identity management, and by a lack of data management that makes fraud particularly difficult to track.

He also noted a prevalence of people perpetrating fraud in Nigeria and escaping abroad with the proceeds. “Fraud is a universal problem that we [governments] need to work together on, to have a common database to be able to fight fraud on a global scale, rather than limiting the problem to country by country,” he said.

Miller gave a number of examples of groups and forums that facilitate cross-border collaboration, including the Five Eyes’ International Public Sector Fraud Forum, the Supreme Audit Institutions (including the UK’s NAO and the US GAO), and the OECD, through its anti-corruption forums.

“But I think we could be doing a lot more. It’s frequently the same few countries that the United States tends to partner with,” she said.

Wolf emphasised that while it’s important to encourage collaboration through international audit networks, “we also need to go beyond that”, echoing Oluata’s point about data sharing.

There is a need to have “data sharing and the interoperability of data systems to break through those barriers” he said. “There’s the opportunity here for alignment around open data standards and shared fraud intelligence to really take it that step further.”  

Read more: Data analytics could prevent billions in fraud, says US oversight committee

The promise and threat of AI

Talk turned to the use of AI by criminals and state actors to perpetrate fraud and the opportunity for governments to use those same AI tools to combat it.

“We’re seeing a lot more fraud happening because [criminals and bad actors] have access to these tools,” Miller said. “So they’re able to craft more believable phishing emails, they’re able to use voice spoofing, they’re able to use deep fakes – they’re able to use all sorts of new technologies that enable them to commit fraud in a far better, more effective and faster way,” she said.

She noted that identity theft-based fraud in particular is growing in the US and that it can be very difficult to discern whether a person is who they say they are, but that the government is increasingly using GenAI for due diligence in this area.

In response to an uptick in fraud whereby perpetrators were stealing identities and then calling the US Social Security Administration (SSA) to change bank account information, the SSA has adopted what’s known as ‘pindrop’ technology, which helps to identify when someone is using a spoofed phone number or if the location of the phone is different to where the caller says they are, Miller explained.

“There are some really promising tools and technologies the government is starting to use to identify fraud in these programmes”, she said, though she also highlighted that it doesn’t have the technological resources of the private sector, which “significantly limits our ability to fight fraud”.

Wolf agreed that AI can help support prevention as well as recovery, but added that “the best outcomes come whenever we have technology, data and people that are aligned around a single source of truth”.

“If we put the technology first without fixing the underlying process, the underlying issue organisationally, data sources, everything like that, the AI is not going to solve all of those problems for you,” he said.

Read more: AI to ‘dominate next four to five years’ of fraud landscape: UK fraud minister

Simpler tech, evaluation, and legal limitations

Reddaway narrowed in on data analytics which he described as “definitely part of the solution”, but he said that while there are “pockets of really good practice” in the UK government, it isn’t widespread and tends to be reactive rather than preventative, primarily due to integration issues.

He emphasised that governments don’t necessarily need AI or sophisticated tech to fight fraud, giving the example of banks, which protect against authorised push payment fraud through simple checks such as linking account name to account number. Yet he said the NAO has found only one UK government body using that same technology.

“There are some basics that can be done,” he said.

He also stressed his concern that people are “pushing tech without pushing good governance”, and that the only way to get fraud controls right is to take an iterative approach.

The NAO advocates – including in international policy forums – what it calls the ‘fraud management cycle’, which is “basically good risk assessment”, according to Reddaway.

He explained: “Really think through how might you be attacked, good controls against those risks, broad loss measurement, and measuring whether or not those controls are working. Evaluate where you are and start again. If you keep doing that, you will refine your controls over time.

“If you try to introduce solutions without having that cycle, you won’t actually understand whether they’re working or not, and frankly, you won’t get the organisational approval to use what can be quite expensive as well. So I would say get the governance right first.”

The ‘Tackling fraud and error across the entire public sector’ webinar was held by Global Government Forum in partnership with knowledge partner Unit4 on 9 October 2025. Watch the webinar in full here and hear the panellists’ answers to a range of other fraud-related questions.