AML Screening with Summy AI
Empower AML match review with LLM.
Sumsub offers an LLM (Large Language Models) review of AML matches as a capability of Summy Copilot, an AI-powered assistant designed to streamline compliance workflows and accelerate various verification checks.
Summy AI-Copilot allows you to:
- Analyze the context of potential AML matches.
- Identify inconsistencies between applicant data and matched record data.
- Generate reasoning notes that explain match classification.
- Provide additional input to Sumsub’s decisioning framework, which determines the final match state.
How Summy AI Copilot works in AML matches review
The review is displayed next to each AML match (the Reviewed by Summy AI button) and opens on-click in a separate drawer.
The review contains the following:
| Section | Description |
|---|---|
| Summary | A short overview of the match, its sources, and the screening verdict. |
| Name analysis | Name matching results. |
| Factual analysis | The comparison of the applicant’s personal info, such as gender and other facts, and the match. |
| Listed records | The source registry entry, its creation date, description, and an age-alignment note. |
| Assessment | A brief explanation of why the match got the suggested state (potential match / true positive / false positive). |
AML Screening with Summy AI FAQ
Can clients disable LLM support if their policy prohibits LLM usage?
Yes. Clients can opt out at any time by submitting a request to Sumsub.
Can clients view the LLM reasoning?
Yes. For each match, the LLM feature generates a detailed explanation of the analysis. You can view it in the automated-review panel attached to the match in the AML Screening results. This explanation is kept separate from the moderator note, which is reserved for manual reviewers.
However, LLM models do not make final decisions. They provide supplemental analysis, which can be used to inform Sumsub’s existing decisioning tools or to supplement manual review.
Which AI models does Sumsub use?
Sumsub uses a combination of advanced models provided by third parties listed in Sumsub’s sub-processors list. The exact model depends on the task and may include sub-pipelines and local models.
Does Sumsub charge an additional cost for this feature?
No. Sumsub does not charge any additional fees for this feature.
Does this feature work with external providers?
No. At the moment, this feature works only with the default ComplyAdvantage integration. Clients who use World Check 1, Quantifind, or ComplyAdvantage Mesh cannot use Sumsub’s LLM analysis feature.
Updated about 23 hours ago