Conceptual illustration – AI and human investigators work together to detect fraud, recover stolen public money and protect the payments people depend on.
An American Treasury success story shows how artificial intelligence, working with human investigators, can help recover stolen public money and strengthen the protection of everyday payments.
By: Ya Libnan – Article 16 in our series about how AI is helping humanity:Analysis
A stolen check is more than a financial transaction. For a retiree, it may mean anxiety about paying the rent. For a family awaiting a tax refund, it may mean postponing an essential purchase. For taxpayers, it means public money diverted from its intended purpose.
That is why one documented achievement deserves a place in the debate about artificial intelligence: the U.S. Treasury reported that AI-assisted check-fraud detection helped recover $1 billion in fiscal year 2024.
A billion dollars recovered. A practical result worth paying attention to.
A result we can measure
In its October 17, 2024 announcement, Treasury said machine learning AI had accelerated the identification of Treasury check fraud, resulting in $1 billion in recovery.
The department’s wider efforts prevented or recovered more than $4 billion in fraud and improper payments that fiscal year. Those efforts included risk screening, prioritizing high-risk transactions and changes to payment processing.
We should be precise: the entire $4 billion was not attributed to AI. The specifically identified AI-assisted check-fraud recovery was $1 billion. That figure is impressive enough without exaggeration.
Source: U.S. Treasury, October 17, 2024.
Giving investigators a faster way to act
The foundation had already been laid. In February 2024, Treasury reported recovering more than $375 million during fiscal year 2023 through an enhanced fraud detection process using AI.
Treasury described a process that helped address check fraud in near real time and accelerated recovery efforts involving financial institutions. It also reported that the enhanced process, together with cooperation with federal law enforcement, had led to active cases and arrests.
These are separate fiscal-year results, not competing estimates of the same recovery.
Source: U.S. Treasury, February 28, 2024.
The practical lesson is straightforward. When suspicious payments can be identified sooner, investigators have a better opportunity to act. Technology can help people focus their attention where it is most urgently needed.
The achievement belongs to a partnership: technology supporting the officials, financial institutions and investigators responsible for pursuing recovery.
Protecting money means protecting people
The public debate about AI often feels distant from ordinary life. People hear about powerful models, enormous investments and predictions about the future. They reasonably ask: What does any of this do for me?
Fraud detection offers a concrete answer.
We want our payments to reach the right people. We want public institutions to safeguard the money entrusted to them. We want investigators equipped to confront financial crime.
Recovering public money supports those goals. It does not mean every fraud victim was immediately reimbursed, or that fraud has been eliminated. It means an important tool has demonstrated a measurable contribution to recovery.
For readers of this series, that distinction matters. Confidence in AI should grow from evidence we can examine.
Human judgment remains essential
A suspicious transaction is a reason to investigate. It is not, by itself, proof that someone committed a crime.
Any system protecting payments must also protect legitimate recipients. People need a way to correct mistakes and challenge decisions. Human officials must remain responsible for reviewing evidence and deciding what happens next.
That is the kind of AI partnership we should encourage: faster detection, careful investigation and clear accountability.
A recovered billion dollars gives us a compelling reason to develop this work further. Public agencies and financial institutions should measure results, explain limitations and use technology to strengthen the people doing the job.
AI is helping humanity when it helps protect the money people depend on—and helps recover the public funds that belong to all of us.
Other Ya Libnan articles in the series
When disaster strikes, AI helps rescuers find where help is needed most
AI is helping save millions of liters of water every day
Six weeks, remarkable progress: How AI and teachers helped students learn in Nigeria
AI is helping scientists see Pancreatic cancer before it becomes visible
AI can help invent tomorrow’s medicines—and build pharmaceutical independence
AI is helping find the people the world’s deadliest infectious disease leaves behind
AI helped doctors find 29% more breast cancers—In a trial of over 105,000 women
AI helped a man with ALS speak nearly two million words
When every minute matters: How AI is helping doctors save Sepsis patients
AI is helping save eyesight before it is lost
AI is helping scientists invent new weapons against superbugs
AI has given scientists a new map of life. Analysis
When AI gives families time to escape a flood
Two patients show how AI can help save precious time
How Artificial Intelligence is already improving our lives, Analysis
