From Siloed Defenses to Trusted AI and Collaboration: Key Takeaways from Sibos 2025
- Quentin Felice
- Oct 6
- 3 min read
Updated: Oct 7
At Sibos 2025, one theme dominated the stage: fraud is evolving faster than our defenses.
Financial institutions are facing an unprecedented challenge. Fraudsters are weaponizing AI, deepfakes, and synthetic identities to launch highly sophisticated scams across channels and borders. Real-time payments add another layer of risk: funds disappear in seconds, leaving little time to react.
The numbers are staggering. In countries with populations as small as 6 million, annual fraud losses already run into the billions. Across the globe, financial institutions are under pressure to stop this escalation but the old, siloed approach is no longer enough.

AI: A Double-Edged Sword
The Sibos 2025 panel we participated to alongside Swift, Deutsche Bank, The Monetary Authority of Singapore and FFIS highlighted a paradox: AI is both the weapon of choice for fraudsters and the most promising defense for the industry.
On one side, criminals use AI to scale scams, impersonate executives, and trick customers into transferring funds. On the other, financial institutions are deploying AI to detect anomalies, automate investigations, and improve fraud prevention.
But AI is only as strong as the data it learns from. And right now, data remains fragmented. No single institution has a complete view of fraud patterns, due to a lack of streamlined access to relevant internal data but also to a diversity of external sources.

Collaboration: The Only Way Forward
Fraud does not respect borders — neither should defenses.
The message from Sibos was clear: collaboration is no longer optional. Banks, regulators, payment providers, and law enforcement need to share intelligence seamlessly if they want to stay ahead.
Momentum is building. From the EU to Singapore, Canada to Australia, new legislation is enabling information sharing across institutions and jurisdictions. But as the panel pointed out, the missing piece is global standards — a trusted operational framework to bring best practices together.
Privacy-Preserving Technologies Change the Game
Historically, collaboration has been held back by one critical concern: data privacy. Institutions feared that sharing customer or transaction data would conflict with regulation, security, or client trust.
That barrier is now being dismantled by privacy-preserving technologies (PETs) such as confidential computing and federated learning. A landmark example came from a SWIFT-led initiative with 13 global banks. Using federated learning, the banks trained stronger fraud detection models together without ever sharing raw data.
This was not a theoretical exercise. It proved that cross-border, cross-institution collaboration can be both effective and compliant. Datavillage was proud to support the technology stack that made this possible.

From Pilot to Production
The breakthrough has been achieved. But as several panelists stressed, the real challenge begins now: scaling collaboration from pilots into daily operations.
That means addressing governance, liability, incentives, and trust frameworks. It means embedding collaborative AI models into existing fraud and AML systems, so they deliver impact in real investigations.
The urgency is clear. Fraudsters are not waiting.
Our Perspective
At Datavillage, we believe the fight against fraud will be won by trusted AI agents that connect institutions without exposing sensitive data.
The Sibos 2025 panel confirmed what we see every day:
Analysts are overwhelmed by alerts.
Institutions struggle with fragmented systems.
Collaboration is essential, but it must be secure, privacy-preserving, and scalable.
By combining federated learning, privacy-enhancing technologies, and secure governance frameworks, we can enable financial institutions to:
Investigate fraud up to 70% faster
Share intelligence without breaching confidentiality
Build resilience against increasingly sophisticated criminal networks
The industry has reached a turning point. The technology exists. The consensus exists. Now it’s time to turn pilots into production, using AI agents to streamline key banks and Financial Institutions processes and make collaboration the standard, not the exception.