UPI is credit-push by design

In a credit-push system, the payer’s side initiates: money leaves my bank because I told my bank to send it. In a debit-pull system, the payee’s side initiates: money leaves my account because someone presented my credentials and asked for it. Cards are pull. UPI’s default flow is push — even the “collect request” is just a message asking me to push.

This one design choice does enormous quiet work. Push systems don’t need the payee to ever hold the payer’s credentials, so there is no PAN-shaped honeypot to breach. Fraud migrates from stolen credentials to social engineering — you can’t pull money out of a UPI user; you have to talk them into pushing it. That is exactly what the fraud statistics show.

For agents, the push/pull distinction becomes the whole ballgame. An agent that can be pulled from is a standing vulnerability. An agent that can only push within a mandate is an auditable actor. Rails that get this backwards will teach everyone why, expensively.