ChainIT proposes a protocol for verifiable authority checks on agent payments

The authors argue that payment systems authenticate users, devices and credentials but do not give a relying party compact, independently verifiable evidence that the presenter of a transaction holds current organizational authority for that exact transaction. They say an agent can produce a valid signature while acting outside its delegated purpose, after its sponsor's authority has changed, or on details a human did not approve.

The protocol separates durable identity, a versioned authority record, sourced evidence, deterministic policy predicates and organization-configured dispositions. Each decision and human approval is bound to a canonical transaction digest, capacity is reserved atomically so limits hold under concurrency, and commit is allowed only through a single-use, sender-constrained credential consumed at the rail adapter.

For AI agent subjects, an Agent Subject Profile adds workload registration, an attested key hierarchy and a sovereignty boundary. The paper specifies enforcement profiles for ACH, wire, card, instant payment and stablecoin rails, and it states residual risks and non-goals. It is ChainIT white paper CIT-PA-WP-7.0 and has a companion paper, Provable Compliance.

Developers building agents that move money may need to check authority at the moment of execution, not only at sign-in. The paper proposes binding each approval to the exact transaction details and making execution credentials single-use, so a changed sponsor authority or altered transaction would not go through on a still-valid signature.

The abstract describes a protocol specification rather than a released library or product, so the design is something to evaluate rather than adopt directly.