Reflections from the OECD Global Anti Corruption & Integrity Forum (GACIF), 26 March 2026

Fraud prevention rarely makes headlines when it works — and that is precisely why it has begun to matter more than ever.
That quiet importance was unmistakable during the final session of the OECD Global Anti Corruption & Integrity Forum (GACIF), held on 26 March 2026 in Paris, where Mens Veritas was present. The session, titled “Digital tools for fraud prevention: Leveraging AI and emerging technologies to safeguard public funds”, brought together senior figures from public oversight bodies, supreme audit institutions, civil society, and the private sector.
Moderated by János Bertok, OECD Deputy Director for Public Governance, the discussion featured Eveline Brito (Office of the Comptroller General of Brazil), Seung Pil Choi (Board of Audit and Inspection of the Republic of Korea), Luiz A. Santos, CFE, PMP (U.S. Inspector General community), Cheri-Leigh Erasmus (Accountability Lab), and Indrani Franchini (private sector). The diversity of institutional roles and country contexts was striking—but what stood out most was how closely their perspectives aligned.
The conversation was not really about technology anymore. Artificial intelligence and data analytics were treated as given. The more difficult questions lay elsewhere: what kind of public institutions are needed when fraud prevention shifts from reacting to wrongdoing to anticipating it? And what happens to trust, accountability, and fairness when prevention becomes the guiding logic?

Fraud Has Changed—and So Has the Risk Landscape

Early in the discussion, the scale of the problem was set out bluntly. Drawing on OECD analysis, János Bertók reminded the room that fraud is no longer a marginal leakage in public finances.

Fraud is not only increasing — it has been identified as the fastest growing risk area

He added a figure that framed the rest of the conversation: just five percent of organizational spending lost to fraud amounts to more than five trillion US dollars annually.
To put this into perspective, Germany’s nominal GDP is just over USD 5 trillion.These numbers matter not simply because they are large, but because they change what fraud represents. At this magnitude, fraud is no longer just a compliance or enforcement issue. It becomes a structural challenge to fiscal credibility and institutional legitimacy.

Just as importantly, fraud today looks very different from the version many public systems were built to confront. Speakers repeatedly described fraud as digital, coordinated, and often transnational—designed to exploit fast moving payment systems, fragmented data, and uneven oversight.

That reality was illustrated vividly by Luiz Santos, reflecting on the oversight of emergency programmes during the COVID 19 pandemic in the United States. Trillions of dollars had to be deployed rapidly to protect livelihoods. Fraud followed the same logic of speed and scale.

Fraud has become much more organized, much more digital, and importantly, much more transnational

The implication was clear. Control systems designed for slow, localised misconduct struggle against adversaries that operate across borders and adapt in real time. Delay itself has become a vulnerability.

From Detecting Fraud to Preventing It

One of the most consequential shifts running through the session was the move away from detecting fraud after losses occur toward preventing fraud before damage is done.
Detection is familiar territory. It fits established accountability frameworks: audit, investigation, sanctions, recovery. Prevention is different. It depends on patterns, signals, and risk indicators—on acting before certainty exists.

This shift featured prominently in the experience shared by Eveline Brito, Deputy Minister at Brazil’s Office of the Comptroller General. She described how Brazil connects citizen complaints, transparency mechanisms, audits, and sanctioning powers into an integrated system, with analytics guiding where attention should be focused.
The goal, she explained, is not automation for its own sake, but earlier, better informed intervention.

AI is fundamental because it generates evidence so that we may choose what to look at and act preventively, before damage is created

That framing captures the deeper change underway. Acting preventively means making judgment calls under uncertainty. It means deciding when risk justifies intervention, and how to act proportionately before wrongdoing is fully established. Those decisions are inherently institutional—and political.

Fraud prevention, in this sense, is no longer just a technical capability. It is a governance responsibility.

The Quiet Constraint: Data Governance

Despite the prominence of artificial intelligence in the session’s title, speakers converged on a more prosaic truth: technology is rarely the main constraint. Data is.
Advanced analytics are only as reliable as the information they process. In many public systems, data is still fragmented across institutions, collected under different standards, and governed by overlapping or unclear rules. When such data is analysed at scale, errors, gaps, and biases tend to grow rather than disappear.

This challenge was addressed from a supreme audit perspective by Seung Pil Choi, Commissioner at Korea’s Board of Audit and Inspection. Korea’s long term investment in digital government has enabled centralized audit systems that draw directly on administrative data, reducing reporting burdens and improving analytical reach. Yet even there, limits remain.
Systems built primarily on past fraud scenarios struggle to keep pace with new types of misconduct. As fraud adapts, audit systems must shift toward identifying anomalous patterns—not only known risks. That, in turn, raises the stakes of data governance and cyber security. In highly digital environments, breaches or misuse of data can rapidly undermine public confidence.

Across different country contexts, a common insight emerged: AI amplifies what is already there. Strong data governance strengthens integrity. Weak governance exposes fragility.

Transparency Without Consequences Is Not Enough

Transparency featured prominently in the discussion—but not as an unqualified solution. Several speakers cautioned that transparency without visible consequences does little to build trust.

From the civil society perspective, Cheri Leigh Erasmus captured this tension directly.

If data and flags sit just within the closed loop of government, we miss an opportunity to strengthen accountability

The issue is not whether data exists, but whether it leads to action that citizens can see and understand. Highly sophisticated internal systems may remain invisible—or appear threatening—if their outputs are not translated into meaningful outcomes.

In contexts where trust in institutions is already fragile, the perception of closed, algorithm informed decision making can erode confidence rather than restore it. This is where civil society organizations, journalists, and oversight actors play a crucial role: turning technical signals into social meaning, and ensuring that early detection translates into reform.

Transparency, the discussion suggested, must be paired with consequence to be credible.

Human Judgment Has Not Been Replaced—It Has Become More Visible

As analytical tools grow more powerful, a familiar concern arises: will machines replace human decision making? The session offered a clear answer. Human judgment has not disappeared; it has become more exposed.
From the private sector perspective, Indrani Franchini emphasized that analytics can surface patterns and risks at scale, but responsibility for interpreting and acting on those insights remains firmly human.

AI is as powerful as the data we feed into it — but humans still need to be in the loop

Algorithms cannot assess proportionality, fairness, or broader social impact. Those assessments remain institutional choices—and they become more visible when decisions are informed by data driven systems.

Public officials are no longer primarily searching for problems; they are deciding how to respond to signals identified by machines. That raises the bar for ethical clarity, skills, and communication. Discretion has not vanished; it has become more accountable.

Rethinking What Success Looks Like

As prevention becomes central, traditional measures of success begin to falter. Recovering funds after fraud has occurred remains important, but it captures only part of the value—and often at the highest cost.

Prevention changes the evaluation challenge. When systems work well, losses do not materialize. Networks are disrupted early. Benefits are real but largely invisible.

Several speakers noted the need for new ways of assessing impact. In Korea, for example, audit institutions are beginning to look beyond financial recovery to consider broader social outcomes. This shift acknowledges that preventing fraud protects not only budgets, but confidence in public institutions.

The challenge is as much political as technical. Prevention requires sustained investment and institutional change, yet its success can be difficult to demonstrate. Without credible narratives about avoided harm, effective systems risk being undervalued.

Readiness Matters More Than Adoption

One conclusion resonated clearly across the session: using advanced analytics in fraud prevention is no longer optional. Fraudsters are already exploiting digital tools. Standing still is not a viable strategy.
Yet there was equal agreement that rapid adoption without readiness carries significant risk. Readiness involves legal clarity, ethical safeguards, sound data governance, skilled personnel, and the ability to explain decisions in plain language.

As Luiz Santos put it succinctly:

The best investigation we conduct is the one we don’t have to conduct

Achieving that outcome, however, requires institutions capable of governing anticipation rather than merely reacting to failure. Technology accelerates consequences—for better or worse.

Integrity in the Age of Anticipation

The closing session of the OECD Global Anti Corruption & Integrity Forum did not aim to offer easy answers. What it revealed instead was a growing convergence across regions and sectors: fraud prevention is becoming an anticipatory function of the state, with direct implications for trust, fairness, and legitimacy.
Artificial intelligence is part of this transformation—but it is not its driver. The decisive factor lies in how institutions choose to govern uncertainty, act on early signals, and demonstrate accountability.

If fraud prevention rarely makes headlines when it works, then the challenge for public institutions today is not only to prevent loss, but to govern anticipation—and to earn trust even when success leaves nothing dramatic to see.