In other words, CBDCs.

> “We trained a foundation model on all of the Stripe historical payment data,” Collison explains. This model is not an off-the-shelf integration with the likes of OpenAI or Anthropic, but Stripe’s own, leveraging open-source infrastructure but fully trained in-house. The targeted outcome: smarter fraud prevention, higher conversion rates, and the ability to make real-time, sophisticated inferences across massive transaction data sets.