According to TechFlow, on November 19, Bloomberg reported that the Monetary Authority of Singapore (MAS) warned that as AI agents transition from "analyzing information" to "directly placing orders," capital may shift rapidly between institutions due to minimal differences in returns or incentives, creating new financial stability risks for liquidity and collateral management.
The MAS assesses that agents can continuously optimize and execute trades around the clock, at speeds far exceeding manual processes, leading to fiercer competition among financial institutions in pricing and liquidity provision; once multiple agents react similarly to the same signals, synchronized capital flows could be amplified.
Regulators have already taken corresponding actions. The MAS and industry partners previously published a white paper titled "Runtime Agent-Based Financial Safeguards" (SAFR) and open-sourced a reference implementation through the Future Financial Institute (FFI); during the Fintech Festival on November 18, the FFI also announced the establishment of an "Agent-Based Monetary Sandbox," enabling financial institutions to test AI agent-initiated payments and treasury management in a controlled environment with predefined authorization, limits, and risk control boundaries. The MAS stressed that institutions must ensure agents operate within user directives, internal policies, and risk limits, with controls encompassing authorization, permissions, identity, accountability, and governance.





