Which banks lead in AI for compliance?

Which banks have AI for compliance on the record, and across how many areas?

The 30 highest of the 139 banks with a compliance use case (tagged aml-kyc, model-risk, governance-general or fair-lending) or a dated AI move on compliance, AML, KYC, sanctions, surveillance, model risk, regulators or audit. Ranked by depth score, then by how many of the four areas the bank covers, then by US AI-index score.

Each row quotes the bank's page on this site; the bank link opens the full record with every source. “Not disclosed” means the page records nothing on that point. Banks absent from the table have no evidence on this topic on their page. The AI index score is the bank's score out of 100 on the US AI index; non-US groups are outside that index and show n/a.

Which compliance areas does each ranked bank cover?

The best stage each ranked bank records in each of the four areas; "none" means its page describes no use case in that area.

How is this ranking computed?

Use cases counted: those tagged aml-kyc, model-risk, governance-general or fair-lending on the bank's page. Moves counted: dated timeline entries that mention compliance, AML, KYC, sanctions, surveillance, model risk, regulators or audit and also state AI (entries that deny AI use, and penalty or enforcement entries, are left out). Depth score = 3 × in-house compliance use cases in production + 2 × in-house ones at other stages + 1 × use cases that are a third-party platform or a filing disclosure (whatever their stage) + 1 per dated compliance AI move + 2 per disclosed compliance AI figure (alerts, screening, false positives). Depth is the first key; breadth (how many of the four areas have at least one use case) is the second and only separates banks with equal depth; the US AI-index score breaks remaining ties, non-US groups last. The stage shown is always the use case's own stage, never inferred from an enforcement order or remediation programme.

Governance use cases here include every governance-general and model-risk entry, not only those naming a responsible-AI structure; the responsible-AI ranking applies that narrower test.

What are the caveats?

Model risk and AI governance entries describe controls over AI as well as AI used for compliance; both are counted because both are what banks disclose under compliance. The AML column is the narrower read of AI doing compliance work.

Fair-lending AI is rarely disclosed (a handful of banks), so few rows cover all four areas. Depth favours banks with long, well-sourced pages; vendor platforms and filing disclosures are down-weighted, but a bank that names its vendor still scores above one that names nothing.

How does this differ from the AI index?

The AI index has no compliance component; this page keeps only AML, KYC, model-risk, governance and fair-lending AI and ranks by depth of evidence, then breadth across those four areas. The AI index and its method are on their own page; the technology leaders and generative AI guides are hand-written cross-bank tables with their own sourcing.