Sibos 2026: LSEG's David White on Cutting False Positives in Risk Screening With Active Intelligence

Quick Summary

Ask compliance teams what their number one problem is and the answer is consistent: they want the highest quality decisions at the lowest cost, and the biggest cost in the process is false positives. David White of LSEG, explains at Sibos 2026 how Active Intelligence, built in partnership with AWS, tackles that problem, and why explainability had to be designed in from day one.

What problem was LSEG trying to solve?

White says LSEG went to customers and asked what their number one problem was. The answer: better decisions at a more efficient cost point, and when LSEG looked at where the cost sat, it was very much around false positives. The aim was simple to state, he says, but technically quite a challenge.

Why are false positives such a problem?

False positives represent cost, because customers have to hire analysts to work through them. They also introduce risk. Checking three results to find the right one is fine. Checking 300 takes time and makes mistakes more likely.

What makes Active Intelligence different?

White highlights the "active" part. LSEG still gathers the best data it can on sanctions, PEPs and adverse media, but it also links that to where it sees noise in the system:

• The conventional wisdom was that more data is better, but LSEG's modelling showed that wasn't necessarily true
• Using its large installed base of signals on how customers make decisions, LSEG works out which data improves match rates and which just adds noise
• So rather than simply adding more data, it adds the data it knows will improve the outcome for the customer

What role did AWS play?

Delivering the product meant leaning on LSEG's partnership with AWS, particularly around Amazon Bedrock, since most of these solutions are now powered by AI. White says the collaboration sped up time to market and helped LSEG make full use of the capabilities on offer.

How does LSEG handle explainability and trust?

AI opens up big opportunities, but compliance is a regulated environment. White says explainability was designed into the system from the start, so customers can defend their decisions to the regulator and internally. It also comes back to basics: customers have to trust the data is correct. That sits at the heart of World-Check and Active Intelligence, with work on how data is collected and quality-checked, and clear records of its provenance and timestamp for use later in the process.

FF NEWS TAKE:

The most interesting claim here is that more data isn't always better. Screening vendors have long competed on the size of their databases. LSEG is arguing that the smarter move is using feedback from how customers actually decide to filter out the data that creates noise. For compliance teams drowning in alerts, fewer and better matches, with an audit trail the regulator will accept, is exactly the pitch they want to hear.