Sui proposes a design that enforces spending limits for AI agent payments.

Sui Foundation on October 28 proposed a design that enforces AI agent spending limits at the payment execution stage, so transactions fail if an agent attempts to spend beyond its approved scope. The design grants agents task authority while requiring the underlying system—not the agent's instructions—to enforce the limits, addressing the risk that agents may misinterpret directives or be swayed by external information. The foundation illustrated the concept with an example of an AI agent purchasing research services, where a user sets a budget usable only in approved apps, with spending caps and usage restrictions. The proposal matters because agentic payments are emerging as a key crypto use case, and enforcement at the protocol level rather than prompt level could become a standard requirement for autonomous commerce. Watch for whether Sui ships this as a native feature or developer toolkit, and whether other chains adopt similar execution-layer spending controls.