Deutsche Bank and Ant Digital Technologies Set AI Guardrails

Deutsche Bank and Ant Digital Technologies published a joint white paper Monday proposing how AI agents could handle selected wholesale-finance tasks while keeping their authority limited and revocable.

The paper maps possible uses across institutional treasury, cross-border payments and trade finance. In treasury, agents could forecast liquidity gaps and assess options for moving funds or executing foreign-exchange trades; for international payments, they could coordinate routes. In trade finance, they could examine documents, flag data discrepancies and prepare payment instructions.

The proposed model, called “Bounded Agentic Execution,” favors controlled actions over unrestricted autonomy. Its reference architecture separates intent and orchestration, control and authorization, and settlement execution. Safeguards include authority limits that can be revoked, identities that can be verified and Know Your Agent (KYA) checks.

Ant Digital Technologies presents its Anvita Flow agent platform as an application example. The platform also featured in an earlier HSBC and Ant Digital AI-agent payment test, which used tokenized bank deposits for real-time settlement on a blockchain testnet. The white paper presents the proposed applications as a way to increase automation and operating efficiency in complex financial operations.