Policy watch: three agent-economy bills moving this quarter

Short briefing on regulatory activity in the US, EU, and UK that actually touches AI agents making transactions.

1. US — Agent Authorization Standard (S.2241)

Senate bill establishing a federal definition of "agent authorization" — the requirement that an automated system acting on a user's behalf have an auditable, revocable delegation chain from a human principal. Bipartisan sponsors. Currently in the Commerce Committee. Meaningful because it would give agent-payment systems (Bippsi included) a federal standard to point at when auditors ask what "authorization" means.

2. EU — Agent Finality amendment to the AI Act

Amendment adding an "agent finality" clause to Article 52 of the AI Act, making agent-initiated transactions legally final unless explicitly disputed by the human principal within 72 hours. Clarifies the dispute timeline for consumer protection. Currently in trilogue. Expected to pass Q3 2026.

3. UK — FCA consultation on machine clients

FCA opened a consultation on whether agents qualify as "clients" under MiFID II for the purpose of investment services. The consultation closes in May. The likely outcome is a new "machine client" sub-category with lighter KYC requirements but stricter audit-trail requirements. Bippsi-style systems fit the proposed definition cleanly.

Why this matters

The agent economy has been running on the assumption that existing consumer-protection frameworks will transfer cleanly. That assumption is now actually being tested. The first two jurisdictions to land cleanly-drafted rules will capture disproportionate share of infrastructure build — the "server rooms in Ireland" effect for a new decade.