A signed reference receipt
The specimen the strip above verifies: a canonical body, its digest, and a signature the public key inside the receipt can check. Synthetic data; no customer evidence.
- Format
governed-receipt/v1Ed25519
AI-native audit & compliance
Evidence.
Under control.
Turn scattered artifacts into a clear, connected audit. AI prepares the evidence. Your team makes the call.
Evidence workbenchIllustration
Original document
Access control policyNorthwind · version 6.0
PolicySource passage§ 2 · Password requirements
All interactive user accounts must use a password of at least 14 characters.
Exact source. Traceable version.
Minimum password length14 characters
Grounded in the document
A stated parameter, with its original passage attached.
Design evidence onlySource → proposal → decisionHuman acceptance required
From source to decision
Connect the evidence.
AI proposes requirement mappings. The source travels with every suggestion.
Workflow illustration
Verify it yourself
A signed synthetic reference specimen — no customer evidence or production custody. Your browser verifies the actual signature with no account or network call.
—
governed-receipt/v1 · 2026-08-24Proof, not promises
Each card is a real object with its own record. The counts are read from the source at build time, never typed.
Receipt Verifier The offline verifier
One HTML file that recomputes a receipt's chain and signature on the reader's own machine. Save it once; it needs no account and makes no network call.
- File
offline-verifier.html40 KB · self-contained
Catalog NIST SP 800-53 Rev. 5, release 5.2.0
One framework release ingested from its issuing source and normalized into a catalog, with the source digest retained so the pin can be re-checked.
- Source
01f37cf90e…9bc062NIST OSCAL Content v1.5.0 · OSCAL 1.2.2
Registry 84 tracked frameworks
Every regime the product tracks, with its issuing authority, current edition, and product standing. Only a regime with a catalog behind it is counted as one.
- Schema
controlframe.framework-registry.v3generated 2026-08-30
Built for accountability
Every step accountable.
AI prepares the work. Your team controls what moves forward.
Your judgment is the boundary.
A named reviewer decides whether the proposed action may proceed. AI cannot approve its own work.
Illustrated workflow · no live run or approval
Explore the safeguards
- Control
- A rule enforced at a specific point in the workflow.
- Gate
- A check that holds work until its condition is met.
- Authority
- The person or policy authorized to let work proceed.
- Receipt
- A linked record of what happened, signed when eligible.
- Verifier
- A tool that checks the record independently, offline.
| Safeguard | When it applies | Who or what decides | Record retained | If it cannot proceed |
|---|---|---|---|---|
| Held write | An agent proposes a write | A named reviewer | Held-for-review record in the decision passport | Refused, or approved and re-run |
| Artifact intake | A file enters quarantine | Pinned scan policy | Scan record with engine and definitions digests | Held in quarantine |
| Package release | A package is submitted for release | Named human sign-off | Hash-pinned release record | Held, with the blocking gate named |
| Recipient access | A recipient opens a package | A live grant | Grant manifest and activation window | Refused when revoked or expired |
| Runner admission | A private runner asks for work | Control-plane token binding | Worker profile and package digest in the token | Denied |
| Evidence freshness | An artifact ages past policy | Freshness policy | Stale controls listed in the sweep receipt | Recollect at source |
A verified record proves integrity. Evidence sufficiency, approval, package eligibility, and release each require their own checks.
By the numbers
Counts the records can back.
- Frameworks tracked in the registryregistry v3 · generated 2026-08-30
- 84
- Frameworks with a control catalog on diskthe honest denominator for any coverage claim
- 6
- Catalog units in NIST SP 800-53 Rev. 5, release 5.2.01,014 active · 182 withdrawn
- 1,196
- Fields in the reference receipt body2 chain steps · one signature · one public key
- 16
Assurance intelligence · official sources
Standards change. Stay informed.
The changes that affect your audit, with the original sources and practical next steps.
Browse source briefingsFedRAMP 20x Class B and Class C pipeline opens August 31.
The Consolidated Rules for 2026 move FedRAMP toward persistently maintained security-decision records, reusable certification data, and human- and machine-readable packages, with the Class B and Class C pipeline scheduled to open August 31, 2026.
What to do next
Connect each security decision to its measure, validation, independent review, change history, and released package rather than rebuilding a static folder for every agency.
- 02
EU AI Act enforcement is now a staged operating calendar.
Commission enforcement powers apply to provisions already in force, while transparency duties apply from August 2026 and the amended high-risk deadlines move to December 2027 and August 2028 by system class.
- 03
NIST publishes draft guidance for using AI in CSF analysis.
The initial public draft of SP 1353 illustrates AI-assisted CSF analysis, planning, implementation, and monitoring while calling for precautions and continuous evaluation and improvement.
- 04
HITRUST CSF v11.8 refreshes authoritative-source mappings.
Version 11.8 continues requirement-statement consolidation and refreshes mappings including PCI DSS v4.0.1 and the AICPA SOC 2 Trust Services Criteria.
01 / The ControlFrame difference
Assurance infrastructure, not another AI checklist.
The defensible system is the chain around intelligence: source-native methods, bounded execution, human-held authority, and one governed record operators and authorized recipients can inspect from their own view.
Four connected layers · one decision recordMethod Graph
Implemented foundationTurn assurance methodology into governed procedure.
Versioned procedures bind scope, approved sources, tools, budgets, and human gates before work begins.
- Version-pinned procedures
- Source and tool allowlists
- Named authority gates
Governed Agent Runtime
ImplementedGive every agent a bounded job and a visible stop.
Collection and testing operate inside declared authority. Exceptions route to people; agents do not inherit release power.
- Scoped agent authority
- Human-in-loop decisions
- Run receipts and replay
Private Assurance Runtime
Purpose-built model plannedKeep specialist reasoning inside the approved boundary.
Customer-controlled endpoint routing exists today. The purpose-built private assurance model for evidence classification, extraction, and drafting remains in development.
- Private endpoint routing
- Task-bounded reasoning
- Evaluation before activation
Governed Receipt · Evidence Passport
Implemented for eligible releasesMake the decision record portable and inspectable.
The passport is the portable decision record. For eligible releases, its governed-receipt schema binds identity, model trace, human disposition, and chain state for offline verification.
- Source-to-decision lineage
- Human disposition retained
- Offline verification path
Regulatory sourceVersioned methodGoverned workHuman decisionPortable proof
Explore the operating system
Compose a governed workflowOpen Engagement StudioCompose the method, then see how an assurance team carries it into fieldwork.
Continuous assurance system
Keep the evidence current between audits.
Observe approved sources, collect through bounded jobs, preserve artifact identity, propose control mappings, route exceptions, and release only through named human authority. The reference below shows the operating contract; tenant execution stays inside activated workspaces.
Reference operating modelContinuous evidence control plane
reference workflowno tenant write · human release authority
01Approved sources
User journeysApproved browser scenarios
interactiveSystems of recordAPIs, exports, events
source-boundDocument estatePolicies, approvals, attestations
versionedCredentials and supervised MFA remain inside the approved customer boundary.
Agentic evidence operations
Execute the work. Keep the proof.
Every configured job carries its source, expected artifact contract, control context, reviewer state, and release boundary.
01
MonitorFreshness and change signals
observing02
CollectApproved source journeys
04 captures03
SealDigest, context, custody
verified04
MapNative control + reuse candidates
reviewRUN-2048evidence.capture → artifact.digest → source.mapreview required
03Governed outputs
Evidence repository
Artifacts, documents, versions, freshness, owners
governed recordControl graph
Source obligation, control IDs, reviewed projections
traceable reuseAudit package
Completeness, exceptions, approvals, release state
human gatedEvidence reuse remains advisory until an authorized reviewer confirms sufficiency.
Between-audit readiness loop
- 01Observe
- 02Collect
- 03Validate
- 04Map
- 05Review
- 06Release
Each mark names a connector definition in the registry. A connection exists for an organization only once its own credential has validated against the provider.
Recorded product evidence
The product is the decision record.
Follow one synthetic healthcare enrollment journey from browser interaction to captured artifact, mapped control, and an explicit human review boundary.
Product recording · synthetic reference
See the workflow, then inspect its custody trace.
The recording shows the browser journey. The companion trace illustrates capture, artifact identity, and the human-release boundary; it is not a signed custody record.
ready to replayRUN-2048 · separate browser scenario
recorded synthetic journeyreference browser boundary
0:00 / 0:00
Framework modules
One evidence spine. Framework-native modules.
Each tracked regime carries explicit source and product standing. Where a catalog and method are released, one governed cross-reference layer can propose reuse without presuming the target reviewer’s conclusion.
A source-backed learning instrument
See exactly where authority ends and judgment begins.
Explore one real release from issuing source to normalized catalog, retained assessment context, and named human acceptance. Every station carries its own status; the path never fills a missing fact with a claim.
Open the Catalog ObservatoryNIST SP 800-53Rev. 5
Release 5.2.0
AuthorityNIST · Release 5.2.0
Source01f37cf90ea9…6e9bc062
Catalog1,196 catalog units
MethodRetained · inactive
DecisionHuman required
- Catalog units
- 1,196
- Active
- 1,014
- Withdrawn
- 182
- Objective nodes
- 3,715
Operating leverage
Stop rebuilding the audit room.
Whether the audit is SOC 2, HITRUST, PCI DSS, HIPAA, or CMS EDE, the handling lives in the evidence. Adjust the hours-first model to your own program and see the effort returned; then confirm it against observed work in a scoped pilot.
Authority boundary
ControlFrame prepares the record. Independent assessors render the opinion.
Qualified assessors, CPA firms, certification bodies, and other authorized third parties keep their existing authority. ControlFrame makes the evidence easier to inspect; it does not certify the customer.Enter the proof room
In the family
A LockedIn Labs application.
ControlFrame is the control plane beside two sibling properties: the LockedIn Labs FDE platform, which runs engagements, and the FDE Benchmark, which measures the work. ControlFrame is where the controls around that work are held and proved. All three are LockedIn Labs properties; these links go to our own sites. About ControlFrame
Pilot one evidence lane
Bring the workflow that costs you most.
We will map its systems, labor, controls, authority gates, and measurable pilot outcome in one executive working session.