In an interview with Deeptech Times, Pedro Bizarro, chief science officer; and Nuno Sebastiao, CEO of Feedzai, a leading AI platform for financial crime prevention; shared insights on emerging threats, regulatory challenges and the company’s innovative strategies in staying ahead in the fight against fraud.
The conversation highlighted how Feedzai leverages AI to combat financial crime and its vision for global expansion. Below is the curated Q&A based on the discussion that offers a glimpse into the future of AI-driven financial security.
How is AI shaping new threats in financial crime, and what trends should we expect in the coming years?
Pedro Bizarro: AI, particularly GenAI, is lowering the barrier to entry for fraudsters, enabling more sophisticated and scalable attacks. We’re seeing five key trends:
- Lower barrier to entry: Platforms like deepfake tools or unrestricted LLMs) allow fraudsters to create attacks with minimal expertise.
- Scale through automation: Automation enables fraudsters to launch thousands of variations of an attack, optimising for success.
- Increased personalisation: Attacks are becoming hyper-targeted, using data from LinkedIn or Instagram to craft tailored scams. A study by IBM’s X-Force showed that personalised attacks, which once took 16 hours, can now be executed in just 5 minutes.
- New types of attacks: Synthetic fraud, like the US$25 million Zoom call scam where fraudsters used fake identities and voices, is on the rise.
- Increased impact: Scams and phishing are now the largest crime category in many countries, with global scam losses reaching US$1 trillion in 2024.
These trends underscore the urgency for advanced AI defenses to counter rapidly evolving threats.
How does Feedzai stay ahead of these malicious trends?
Pedro Bizarro: We focus on two main strategies: prevention and productivity. For prevention, we develop advanced models that detect fraud by analysing diverse data points—transaction history, card details, device information, GPS and even typing patterns. Our recently launched Scam Alert product uses large language and vision models to analyse phone screenshots, and flag suspicious messages or websites in real time.
On the productivity side, we enhance fraud analysts’ efficiency. Our AI-driven case manager automates 99 per cent of transaction decisions, leaving only a small fraction for human review. To streamline this, we’ve built AI agents that pull and summarise data from sources like Google Maps or Wikipedia, reducing the time analysts spend on manual research. This holistic approach ensures we stay proactive in combating fraud.
IMAGE: Feedzai
Can you explain how Feedzai’s RiskOps framework helps financial institutions tackle these threats?
Pedro Bizarro: Fraudsters don’t operate in silos—they exploit multiple channels, from account openings to money laundering. Traditional financial institutions often have fragmented teams handling different types of fraud, which limits their visibility.
Our RiskOps framework unifies data across the entire customer lifecycle—account opening, transactions, KYC and anti-money laundering. By integrating these data points, we enable better decision-making to allow institutions to track and stop fraudsters as they move across channels. This unified approach mirrors the agility of fraudsters, giving the “good guys” a fighting chance.
What role should regulators play in combating financial crime through AI, and how can they ensure transparency and fairness?
Pedro Bizarro: Regulators have immense potential to drive positive change if they balance enforcement with innovation. A great example is Brazil’s PIX system, a real-time payment platform mandated by the central bank to boost financial inclusion. By requiring banks to adopt PIX while also lowering barriers for new banks to enter the market, regulators created a vibrant ecosystem that increased transparency, reduced cash-based “dark pool” transactions and boosted tax collection. This shows how regulators can enforce rules while fostering innovation.
Contrast this with Europe’s AI Act, which, despite good intentions, is overly complex with ambiguous guidelines. For instance, it bans AI from ‘influencing’ people, but this could inadvertently affect legitimate uses like personalised marketing. Regulators need to simplify frameworks to ensure clarity and compliance without stifling innovation.
What safeguards should financial institutions implement to avoid misuse, bias or overreliance on AI systems?
Pedro Bizarro: The key is to monitor the entire system, not just the AI model but also the rules, user interface and human interactions.
For example, we ran an experiment comparing models with and without AI explanations. While explanations improved transparency, they sometimes led analysts to over-rely on the system, making quicker but less accurate decisions. This taught us that banks must clearly define what they’re optimising for—speed, accuracy or a balance—and continuously measure outcomes. Experimentation, monitoring and adaptability are critical to avoiding bias and ensuring robust AI systems.
Who are some of Feedzai’s key clients, and how do you address regional differences in deploying your technology?
Nuno Sebastiao: We work with major global banks like HSBC and Standard Chartered, as well as regional players like Trust Bank in Singapore. Deploying technology varies significantly by region due to regulatory and cultural differences.
For example, implementing our solutions in the UK differs from that in Hong Kong or Mexico. Our approach is to “think globally, act locally”, which means hiring local talent with deep market knowledge to bridge our technology with regional needs. This ensures our solutions are tailored to local contexts, whether it’s Singapore, Indonesia or Australia, where we recently acquired a company to strengthen our presence.
IMAGE: Feedzai
What are Feedzai’s growth plans, and which regions are you targeting?
Nuno Sebastiao: We’re doubling down on existing markets like Singapore, Hong Kong and Australia, where we’ve tripled our headcount to about 70 people. We’re also exploring opportunities in Indonesia due to strong demand for our solutions, balanced against strict local governance requirements.
Globally, we see greater sophistication and adoption of financial technologies in regions like Asia and Latin America compared to Europe or the US. These markets are leapfrogging older systems, creating opportunities for us to innovate and later bring those advancements to Western markets. China isn’t a focus yet, but we’re open to future expansion with the right flagship clients.
What new innovations are on Feedzai’s product roadmap for the next six months?
Nuno Sebastiao: We’re tackling two major challenges: data ingestion and immediate value delivery.
First, we acquired an orchestration company in Australia to streamline how we integrate diverse data sources across regions, significantly reducing the time from agreement to production.
Second, we’re launching Feedzai Zero, a product that leverages our global payment graph—the largest and most diverse outside Visa and Mastercard. This allows us to provide instant fraud and risk insights in new markets or for new products, eliminating the need for months of data training. These innovations aim to shorten time-to-value from months to days, and tranform how banks assess risk.
How does Feedzai ensure its solutions are production-ready and compliant across diverse markets?
Nuno Sebastiao: Deploying machine learning at scale in critical systems like finance is challenging due to regulation and the need for consistency and explainability.
Our decade-long experience has taught us how to navigate these complexities. We focus on compliance with local regulations, whether in Europe, Asia or elsewhere, while ensuring our systems deliver reliable, actionable results.
By combining deep research, local expertise and robust orchestration, we’ve mastered putting AI into production, setting us apart in the industry.
