Mizuho and NEC pilot KYA to verify banking AI agents’ identities

KYA is an early glimpse of the institutional plumbing we’ll need if AI agents are ever going to move real money at scale. By treating AI systems as first‑class “customers” whose identity needs to be verified—just like humans under KYC rules—Mizuho and NEC are acknowledging that tomorrow’s banking clients may be autonomous software that schedules payments, reallocates portfolios, or negotiates credit on behalf of people and firms. The joint trial is narrow, but conceptually it starts to answer: who vouches for an agent, and how does a bank know which model is actually behind an API call?([itmedia.co.jp](https://www.itmedia.co.jp/news/articles/2605/28/news124.html?utm_source=openai))

Strategically, this is a defensive move against both fraud and platform disintermediation. If banks can establish their own trust‑anchoring schemes for agents, they’re less likely to be forced into whatever identity fabric big tech companies build into their app ecosystems. It also gives regulators a tangible object—agent credentials—to tie emerging rules to, rather than trying to regulate prompts and weights directly.

For the AGI race, KYA doesn’t change raw capability, but it does make a world of AI financial agents slightly more governable. That could, paradoxically, accelerate adoption of highly capable systems in finance, because CIOs and supervisors can point to a concrete control: only authenticated, whitelisted agents get transaction rights.