Citadel AI to Supply AI Governance Solution to Sumitomo Mitsui Banking
A major Japanese bank is buying outside tooling for generative AI risk control rather than building it in-house, which puts a vendor inside the compliance layer of customer-facing and internal AI use.
Reporting from 1 source: ASCII.jp.
Citadel AI will provide an AI governance solution to Sumitomo Mitsui Banking Corporation, the company said. The bank will use it to standardize risk verification for generative AI, run continuous monitoring, and build a common platform with guardrail functions. The system watches prompts and responses in real time, detecting harmful inputs and unsafe outputs, then triggering blocks and alerts. Citadel AI says it will support the bank's company-wide AI governance.
The deal covers three things: a shared set of evaluation metrics and test data for assessing generative AI risk, an operational process for verifying risks before an application goes live, and real-time monitoring of prompts and responses once it is running.
Citadel AI says the monitoring detects inputs that try to extract confidential information and responses that lack safety, and connects those detections to blocking and alert notifications. Sumitomo Mitsui Banking is developing a common platform with guardrail functions, and these functions feed into it.
The stated rationale is that pre-launch checks alone are not enough once AI reaches customer-facing services. For financial institutions, the company cites customer protection, information management, and regulatory compliance as reasons the management framework has to match each use case.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.
Sources
- ASCII.jp Citadel AI、三井住友銀行にAIガバナンスソリューションを提供