AI, Trust, and Collaboration: Rethinking Fraud Prevention in the Age of Open Payments
At the Payments Canada Summit, this panel — moderated by Lisa Ford, CLPO of Payments Canada — brought together perspectives from across the financial-crime stack: a global anti-fraud AI vendor (Feedzai), an open banking infrastructure provider (Banfico), Canada's central bank supervisory function (Bank of Canada), and the federal policy lead (Finance Canada). The discussion explored how artificial intelligence, data collaboration, and trust frameworks are reshaping financial crime prevention and the future of payments — and how regulation, supervision, and industry standards must move in lockstep to enable the next era of open finance and agentic transactions.
The panel composition itself was deliberate: a supervisor, a policy lead, a detection vendor, a trust-infrastructure operator, and a moderator from the rail itself — all in the same conversation. Lisa Ford framed the closing question in a line that captures the panel's intent: "how we all strive to bridge the gap between rapid speed and systemic safety." It is a question no single seat on the panel could answer alone.
Andy Renshaw, SVP Product Management, Feedzai — AI Works, Collaboration Multiplies It
Andy Renshaw of Feedzai opened with a pragmatic view on the current state of AI in financial crime. "Does it work? Does it make a difference? And the answer is yes," he said, noting that most institutions are already seeing "a 25 to kind of 35% improvement in the financial crime prevention." But the real breakthrough comes when institutions move beyond isolated deployments. "When you move into collaborative or network mode, you can probably add another 50% to that again," he explained, pushing the figure to roughly 40–45%. "Collaboration is vital and data is vital within that AI context."
Renshaw also drew a distinction that matters for any executive deploying AI inside a regulated environment. "As consumers, we're all kind of AI native now. The AI in business and financial crime prevention is not quite the same thing." Enterprise-grade AI in financial services, he stressed, has to be engineered for explainability, governance, and operational scale — and that is materially different from the consumer-grade chat experiences shaping public perception.
Drawing on his international experience (Feedzai is headquartered with significant operations in the UK), Renshaw highlighted how different jurisdictions are operationalizing AI in financial crime in materially different ways:
- United Kingdom: Confirmation of Payee and the split-liability model. Whatever one's view on whether split liability is the right policy, "having a liability model and the confirmation of payee model in the UK has created labels." And labels, Renshaw noted, are "really useful as I'm refining your AI decision." A regulatory choice quietly improved data quality for the entire ecosystem — an unintended consequence worth studying.
- Singapore: A central operational anti-scam capability. "They adopted a central kind of operational prevention around scam. And they're using AI to find connections… imagine thousands of calls coming in every day. They're using AI to combine those insights and message back to business."
- Brazil: A market where the pace of adversarial adaptation has forced real-time AI. "They have to use AI within their payments context to adapt what they're doing. What they'd be doing on a Monday isn't working on a Thursday. You've got to change it by the next week. You're not going to do that with some of the legacy kind of approaches."
- The Netherlands: Data pooling and data sharing have "proved the value that more data in the right context and the standard format makes a difference."
- Australia: Pushing further by combining "telco data, social media content, and payment content" — extending visibility beyond the transaction itself to capture the activity that comes before and after.
He closed with an analogy worth quoting: enterprise AI in fraud is like the cars on the road. The car you bought a few months ago was being built two or three years before that. "Once technology kind of depends on the road, I guess that's why I think we'll be looking at it." The implication: innovation must be continuous because the gap between deployed and emerging capability is structurally widening.
Paulo Barbosa, COO, Banfico — Trust, Governance, and Open Frameworks
Paulo Barbosa, COO of Banfico, focused on trust, governance, and the role of collaboration in scaling open finance. Banfico operates a directory and trust framework for open banking ecosystems — exactly the kind of infrastructure that becomes increasingly important as more actors (fintechs, new market entrants, payment initiation services, and eventually AI agents) join the system.
His framing of the trust problem:
"We have our own directory service, our own trust framework, where we can make sure that this umbrella — all the standards that will then govern the implementation of an open banking system — is enforcing that the different players, the new market entrants… fintechs that are not necessarily the same level of regulation as banks or insurers or even open health, that these new stakeholders coming into this framework, they are all abiding by the same rules and that trust is not broken."
Barbosa stressed that the directory and trust framework must be designed to scale. As open banking expands and new categories of participants join the ecosystem, the integrity of the system depends on every participant operating under the same enforceable rules — independent of size, regulation level, or vertical (banking, insurance, even open health).
He also signalled forward-readiness for agent-initiated payments, while being careful not to overclaim. While Banfico does not yet treat agentic payments as imminent, the infrastructure is being designed with that future in mind:
"I just learned that the agentic AI [and] agentic payment story is a bit far away. We are prepared. We are looking at how can our open banking directory and how can our open payments framework make sure that if and when the time is right to implement [agentic] payments, we can support that and make sure that trust is not broken across these different interactions independently of the framework we are talking about."
The goal is to ensure trust persists regardless of how the ecosystem evolves — and regardless of whether the entity on the other side of the rail is a human, a regulated institution, a fintech, or, eventually, an agent.
Closing Take
In response to the moderator's rapid-fire ask for one mindset takeaway, Barbosa returned to the panel's central message:
"Collaboration is key. The regulation is the baseline that we as an industry need to push the ceiling and make sure that we collectively create safe and trusted solutions for consumers."
He explicitly acknowledged Canada's leadership in this kind of cross-industry collaboration, noting that Banfico and others are trying to replicate similar forums in Europe and other parts of the world.
Supervisory and Policy Voices: Anne Butler (Bank of Canada) and Judith Hamel (Finance Canada)
Across the panel, the supervisory and policy seats added the institutional weight that distinguishes a serious risk-resiliency conversation from a purely technical one.
Anne Butler, Managing Director, Supervision, Bank of Canada, represented the central bank's supervisory perspective. Bank of Canada Supervision oversees prudentially regulated payment service providers under the Retail Payments Activities Act (RPAA) — a regime that has materially changed who is accountable for what in the Canadian payments stack. The supervisory question on this panel — what does prudent, risk-aware payments innovation actually look like when AI, scams, and open banking are all evolving at once? — sits squarely in her brief.
Judith Hamel, Director General, Financial Services Division, Department of Finance Canada, brought the federal policy lens. Finance Canada has been the lead department behind Canada's evolving payments policy stack — open banking framework, payment modernization, anti-fraud strategy, and the legislative work that sits beneath them. Her role on this panel underscored that the rules of engagement for AI, agentic payments, and open finance are being actively drafted, not waiting for a future legislative cycle.
The combined presence of these two seats — supervision and policy — made it unmistakable that trust and transparency in payments are no longer purely industry concerns. They are now an explicitly co-governed responsibility shared between rail operators, infrastructure providers, vendors, supervisors, and federal policymakers.
The Bottom Line
The combined message from the panel points to a clear conclusion: AI alone is not enough. Its true value emerges when combined with high-quality labelled data, shared intelligence, and robust trust frameworks. As Canada's payments ecosystem expands to include new participants — fintechs, payment initiation providers, and eventually autonomous agents — the challenge is no longer just detecting fraud. It is coordinating trust across an increasingly diverse and dynamic network of human and machine actors, with both supervision and policy moving in step with the technology.
Renshaw shows that the data-and-network thesis works in practice across multiple jurisdictions. Barbosa shows that the infrastructure to enforce trust across those networks is being built today — and is being designed with agentic payments in mind, even if that future is not yet imminent. Butler and Hamel anchor that work in the supervisory and policy frameworks that will determine whether it scales safely. And Ford, from the seat of Payments Canada, frames the central question: how do we bridge the gap between rapid speed and systemic safety?
Key Takeaways
- 25–35% lift in financial crime prevention from AI at a single institution; +50% more (to roughly 40–45%) in collaborative / network mode.
- Enterprise AI in financial crime ≠ consumer-grade AI — explainability, governance, and operational scale are different problems.
- Global learnings: UK (Confirmation of Payee + split liability inadvertently creating labelled training data), Singapore (centralized scam-call signal aggregation), Brazil (Monday-vs-Thursday adversarial speed forcing real-time AI), Netherlands (data pooling), Australia (cross-channel: telco + social + payments).
- Banfico's directory service and trust framework are positioned as the connective tissue for open banking — and explicitly designed to be ready for agentic payments "if and when the time is right."
- Regulation is the floor, not the ceiling — the industry must collectively push higher.
- Canada is recognized internationally as a leader in cross-industry forums of this type; Banfico is replicating the model in Europe.
- The panel's composition — Bank of Canada (supervision) + Finance Canada (policy) + Banfico (trust infrastructure) + Feedzai (detection) + Payments Canada (moderation) — is itself the message: risk-resiliency is now a four-corner conversation.
