Right to Send Your Agent: Citizen Power in the AI Services Economy

What is broken and why it matters

What we would change

1. 1

   Create a national delegated-agent standard for identity, authentication, authorization scopes, revocation, expiration, receipts, audit logs, liability, incident reporting, and anti-fraud review.

2. 2

   Require federal services and high-impact regulated sectors to accept user-authorized agents through secure APIs where they already must accept a person, legal representative, or authorized third party.

3. 3

   Build a consent firewall for the agent economy: every delegated action should have a stated purpose, data-minimization rule, time limit, receipt, and one-click revocation path.

4. 4

   Replace password sharing and brittle screen scraping with standards-based agent identity, verifiable credentials, least-privilege access, and public conformance tests.

5. 5

   Create a public-interest Life Admin Autopilot, built through legal aid, benefits navigators, consumer-protection agencies, libraries, state digital services, and civic technologists, to help people assemble evidence, track deadlines, renew benefits, compare bills, dispute debt, appeal denials, and prepare forms.

6. 6

   Give every high-impact automated public decision an appeal packet: the rule applied, the evidence used, the deadline clock, the missing documents, and a human escalation route.

7. 7

   Phase the mandate by risk and capacity, beginning with federal services and sectors that already have data-access or interoperability rules, including finance and health care.

8. 8

   Preserve human and paper alternatives so agent access expands power for people who want it without excluding people who cannot or should not use it.