LARRI and governance
Eight institutions describe the same missing record.
Journalists, senators, clinicians, federal auditors, Canadian public-service faculty, and academic researchers have each examined automated decision-making in healthcare and arrived at a problem that sits at the integrity layer. LARRI is designed to address that layer. Hear from officials and researchers, in their own words, about the problems LARRI is built to answer.
Public Broadcasting Servicepbs.org 
Canada School of Public Servicecsps-efpc.gc.ca 
Committee on Financefinance.senate.gov 
Health, Education, Labor & Pensionshelp.senate.gov 
Office of Inspector Generaloig.hhs.gov 
Permanent Subcommittee on Investigationshsgac.senate.gov 
Government Accountability Officegao.gov 
University of California / UCTVuctv.tv Governance and automation
In the age of AI, records need better integrity. Each account explains why
Observed practiceHow algorithms are used to deny health insurance claims
PBS NewsHour
PBS reports that claims could be rejected without opening the patient file, and that medical directors signed batches covering fifty patients with decisions averaging under two seconds. A signature produced in that workflow attaches to a database event. LARRI binds a reviewer's attestation to an exact decision package, which makes the signature attributable to the operative evidence and the policy in force when the determination was made.
Open the original sourceWhere regulation is converging
Governments have already named the required qualities.
Regulators have stopped asking whether automated decisions need governing and started identifying what governance must produce. ONC’s HTI-1 rule demands source attributes and risk management for predictive interventions. CMS demands specific denial reasons, standardized prior-authorization exchange, published metrics, and coverage criteria consistent with Medicare rules. NIST and GAO supply the lifecycle accountability frameworks around them. Every one of these instruments describes a decision an institution must be able to account for after the fact, and none of them specifies the record that would make the accounting possible.
Nine standards, one artifact
How LARRI meets the standards
Institutions are converging on the standards LARRI is built to uphold: every consequential automated decision should carry exact, reviewable evidence of its system, sources, purpose, risks, reasons, monitoring, human responsibility, and governing authority.
LARRI turns institutional governance into one independently verifiable decision artifact, binding every consequential outcome to its exact system, sources, intended use, risk controls, reasons, exchange context, audit history, responsible human action, and governing authority.
LARRI binds each decision to the declared system, policy, configuration, profiles, and verification contracts in force at that moment. The resulting record makes the operative decision context visible without requiring disclosure of proprietary internal logic.
LARRI can commit to source records, datasets, model documentation, evaluation reports, and performance evidence by exact digest and artifact identity. A verifier can confirm that the cited material is the same material originally bound to the decision.
Verification targets, selected profiles, permitted claims, and residual limitations are explicit parts of the public result. Every PASS remains confined to the exact assurance that was selected and verified.
A LARRI profile can require identified risk assessments, controls, evaluations, and approval records as decision evidence. The receipt preserves which risk-management materials governed the event and whether the required public checks passed.
A domain profile can require every adverse decision to carry a structured reason, governing rule, evidence references, and responsible attestation. Those elements become part of the signed decision record rather than a later narrative reconstruction.
LARRI uses exact schemas, canonical encodings, registered identifiers, content-derived identities, and deterministic normalized results. Independent systems can exchange the same artifact and reproduce the same public verification outcome from the same inputs.
Each receipt provides a portable, independently verifiable record of what was declared, committed, signed, and checked. Collections of receipts can support oversight across systems, vendors, policies, and decision classes while preserving the identity of each individual event.
A human action, reviewer role, signer declaration, and attestation can be bound to the exact decision package. Responsibility is therefore attached to a specific record state rather than inferred from a generic workflow log.
LARRI makes governing contracts, trust inputs, profiles, limitations, and conformance evidence explicit and independently inspectable. Institutions can demonstrate which rules governed a decision and outside reviewers can verify that those rules were applied to the declared record.
Normative foundation
Read how the receipt and bundle model works.
The protocol binds assertions, commitments, contracts, signer declarations, exact file bytes, and stated limitations for independent verification.