The Intelligence No Single Bank Can Build Alone

Sibos arrives in Miami next week under the theme of “Digital Finance for AI-Driven Economies.” As a twenty-year-plus veteran of the financial services industry, I’m looking forward to attending and having conversations about what’s next for the industry.

As the conference title suggests, AI will take center stage — how to deploy it, how to govern it, and how to get it to work on data that cannot leave the building. I’d argue that the last point is the one that matters most. The most important conversations next week won't be about the models. They'll be all about who gets to use the data.

The Challenge of AI-Ready Data

Almost none of AI use cases in financial services are solvable by one team, one department, in one institution, operating on one dataset. 

In the past, the default response to the lack of data has been to centralize into one warehouse, one lakehouse, or one place where everything can sit together. For many of the highest-value use cases in financial services, that option simply is not available. Not because the budget isn't there, but because the data cannot move.

The constraints that prevent centralization are structural — GDPR, banking secrecy laws, data residency requirements, and commercial sensitivity. Any AI strategy built on the assumption of consolidation has a ceiling most institutions will hit relatively quickly and long before they reach scale. 

Relocate the Compute Instead of the Data

Federated computing is the answer to the problem posed by centralization’s constraints. Data stays where it is inside its owner's environment and jurisdiction, and code is executed locally. The model, the analysis, and the AI agent all travel to the data. What comes back are model updates, insights, and completed workflows — never the records themselves.

So, what does this actually look like inside a financial institution?

Rhino Data Activation harmonizes siloed data across core banking, payments, risk, and compliance systems into governed, AI-ready infrastructure with a common data schema, without consolidating it or moving it across jurisdictions. For institutions sitting on decades of fragmented proprietary data that their AI initiatives cannot yet touch, this is where internal readiness starts. Rhino Data Activation also allows financial institutions to build better customer experiences, enhance decision-making processes and build new products by stitching together their own siloed data to create new levels of insight.

Rhino Federated Intelligence expands the same principle across institutional lines. Banks and payment networks can collaborate to build shared fraud, AML, and sanctions screening models across distributed transaction data. Or an open, governed data collaboration with fintechs, insurers, and analytics partners in which the participants can safely trust that their intellectual property remains completely under their control thanks to the privacy-enhancing technologies (PETs) built into its infrastructure.

Both solutions are powered by the Rhino Federated Computing Platform (Rhino FCP) that also supports secure and governed agentic workflows. Financial institutions can deploy AI agents across distributed data environments, orchestrate cross-site workflows, and connect existing agentic systems to live data — all with data privacy and permissioning guardrails controlling every step. As internal data becomes harmonized and governed through Rhino Data Activation, it also becomes the foundation financial institutions can build today and take positive steps toward the fully agentic operating model that the future holds.

The federated approach has already been validated at scale in financial services. Working with Google Cloud, Swift ran a federated fraud detection initiative across 13 global financial institutions, with Rhino’s Federated Computing Platform delivering the core federated learning platform. Shared intelligence across institutions substantially outperformed any model a single bank could train on its own data.

J.P. Morgan and BNY reached a similar conclusion through Project Aikya, a federated learning proof-of-concept for anomaly detection in cross-border payments. Models trained collaboratively across institutions outperformed those trained on any single institution's data alone.

Federated Computing at Work: Five Use Cases

The same architecture applies across the problems that have resisted solution precisely because they require data no one institution controls.

Agent Governance and Observability

As agentic AI spreads across the enterprise, most institutions lack visibility into how those agents actually behave or what data they are touching. Federated computing lets organizations monitor and govern every agent across whatever platforms built them, with a trusted framework governing which agents can act on which data.

Observability and Digital Assets

As the industry is moving and quickly adopting blockchain technologies for digital asset services, the same technique can provide observability across nodes in a Distributed Ledger Technology (DLT) network to guarantee smooth operations and resilience. Where financial transactions are executed “on-chain,” much of the data remains siloed “off-chain” which can be unlocked with privacy-safe federated computing.

Liquidity and Treasury Visibility

Cash obligations land across many channels and systems. Without a view spanning all of them, capital sits trapped and intra-day operations run on incomplete information. Federated visibility across systems lowers the cost of carry and takes friction out of treasury operations.

Trade and Supply Chain Finance

Every trade flow involves multiple banks, corporates, and jurisdictions. Collaborating across that chain improves transparency, eases compliance checks, and makes counterparty exposures far more predictable, without requiring any participant to expose their proprietary data.

Cross-Institution Fraud, AML, and Sanctions Screening

Fraud and money laundering operations exploit the visibility gaps between institutions. Federated learning lets banks build shared detection models across distributed transaction data, generating collective intelligence that no single institution could build alone, without a single customer record ever leaving its source.

The common thread across all four patterns is that the highest-value work requires data no single institution controls. Federated computing is what makes that data accessible without exposing it.

Ready to Work Through What's Blocking Your AI Roadmap?

The challenges described here are live on the desks of almost every institution I speak with. If any of them map to where you are, I’m happy to go deeper. I will be at SIBOS September 28 through October 1 — schedule time on my calendar so we can connect in Miami.

Not attending Sibos? Get in touch with the Rhino team so we can start the conversation.