# Open proposal to Congress: auditable decision records for consumer AI agents

Version 1.0  
July 29, 2026

Published as an open proposal on July 29, 2026. It has not yet been submitted to Senator Warner’s office.

## Purpose

Senator Mark Warner’s June 2026 AI AGENT Act discussion draft asks how consumer-facing agents can act in users’ interests while supporting privacy, cybersecurity, interoperability, competition, and accountability. This proposal offers a testable addition to that discussion: an optional, versioned decision-record profile for consequential consumer-agent transactions.

A transaction receipt shows what happened. It usually cannot show which serious alternatives the agent considered, what it predicted, why it rejected other options, which feeds supplied the evidence, or whether the result later matched the user’s instruction. A decision record preserves those objects so an authorized evaluator can test specific loyalty, steering, and performance questions.

## Proposed congressional action

Congress should invite NIST, the Federal Trade Commission, consumer representatives, agent providers, marketplaces, independent researchers, and privacy and security specialists to evaluate a narrowly scoped pilot for the Decision Record & Control Standard Consumer E-Commerce profile (DRCS-EC).

The pilot should:

1. define a common record shape for a bounded set of consequential consumer-agent transactions;
2. test whether independent implementations can produce and validate the same required objects;
3. minimize retained personal and commercial data;
4. separate record production from the rules, readers, and consequences that may use it;
5. publish conformance fixtures, threat models, failure reports, and version changes; and
6. measure whether the profile helps evaluators detect steering, instruction drift, omitted serious alternatives, or poor follow-up without treating the record as proof by itself.

The pilot would supply evidence for later legislative decisions. It would not predetermine a mandate, an enforcement rule, or a finding about any provider.

## Minimum consumer-commerce record

The DRCS-EC 1.0 profile proposes:

- a delegation snapshot stating the user’s task, constraints, authority, and consent;
- an option-set snapshot identifying the serious products, sellers, prices, shipping terms, return terms, payment paths, checkout friction, and reliability information considered;
- the selected option and the reasons for selection;
- rejected serious alternatives and the reasons they were rejected;
- a scoring trace that can be recomputed from the recorded inputs and declared rule;
- feed and evidence provenance;
- outcome follow-up, including whether the transaction satisfied the instruction and what changed after the decision;
- schema version, record identity, timestamps, and integrity controls; and
- the relevant fields of the 17-field DRCS Core record, including evidence, authority, uncertainty, prediction, alternatives, stakeholders, system welfare, review conditions, and reversal triggers.

Record minimization is part of the design. The profile should require only the data needed for the declared evaluation and should specify retention, access, correction, deletion, and disclosure rules.

## What the current experiment supports

The July 2026 working paper *Standardize the Record, Not the Reader* uses a corrected implementation of the Calvano algorithmic-pricing environment to separate several mechanisms.

- Producing the structured evidence path alone leaves pricing unchanged under common seeds.
- Forecast scoring leaves pricing unchanged in this implementation.
- Changing the private reward changes the agent’s objective and conduct.
- Outcome-contingent restrictions change conduct at the higher audit intensities tested, where targeted restrictions outperform equally frequent random interruption.
- The rejected-alternatives addition produces no incremental effect because its trigger is inactive or creates negligible exposure.
- The implemented several-reader composite also adds no consequence because it never activates.

These results support one design boundary: a record is evidence. The experiment does not establish that durable filing works in deployed consumer markets, that DRCS-EC improves field outcomes, that one reader institution is best, or that production cost is known. Those are pilot questions.

## Conformance and evaluation

A useful pilot should test:

- required-object completeness;
- schema and policy version identity;
- pre-decision timestamp and append-only amendment history;
- option-set and feed provenance;
- scoring-trace reproducibility;
- detection of a dominating rejected option under the declared evaluation rule;
- authorization and purpose limitation for readers;
- privacy minimization and data-subject correction paths;
- export and interoperability across providers;
- tamper evidence and record-link integrity;
- false-positive and false-negative rates for each evaluator rule; and
- public reporting of inactive triggers, failed hypotheses, and boundary conditions.

No record should be treated as self-authenticating. An evaluator must still examine provenance, implementation integrity, the declared scoring rule, material omissions, and the institution’s legal authority.

## Open questions

The pilot should invite public answers to these questions:

1. Which consumer-agent decisions are consequential enough to justify a record?
2. Which option-set and provenance fields are necessary for loyalty and steering tests?
3. What information can be minimized, aggregated, held locally, or disclosed only under authorized review?
4. Which conformance tests are technical, and which require a legal or institutional judgment?
5. How should an evaluator distinguish a justified rejection from steering?
6. Which appeal, correction, and supersession paths should apply?
7. What evidence would justify expanding, narrowing, or ending the pilot?

## Sources and downloads

- [AI AGENT Act discussion draft and public summary](https://www.warner.senate.gov/newsroom/press-releases/warner-unveils-discussion-draft-of-legislation-to-create-innovative-market-for-secure-artificial-intelligence-agents/)
- [Calvano/DRCS working paper, figures, methods, and downloads](https://decisionaccounting.org/calvano-drcs/)
- [Calvano/DRCS v1.2 replication package](https://decisionaccounting.org/research/calvano-drcs/calvano-drcs-v1.2-replication-package.zip)
- [DRCS-EC 1.0 technical profile](https://decisionaccounting.org/research/calvano-drcs/DRCS-EC-v1.0.md)
- [DRCS Core 17-field public record guide](https://decisionaccounting.org/templates/)
- [Decision Accounting Academy](https://decisionaccounting.org/academy/)
- [Theorem Laboratory](https://decisionaccounting.org/theorem-laboratory/)

## Corrections and supersession

This is version 1.0. Corrections will be listed on the public proposal page and in the site corrections log. Any successor will identify the exact version it replaces, preserve this file, explain material changes, and link to the superseding version.

Feedback: https://decisionaccounting.org/feedback/?topic=congress-open-proposal