Name, version, and deployment context
FREE CODE STARTER · SOURCE-CHECKED 18 JULY 2026
Make the evidence
reviewable before launch.
Keep AI transparency control claims, owners, test dates, and evidence references in one machine-checkable internal record. The starter includes a JSON Schema, fictional sample, dependency-free Node validator, and six deterministic tests.
No signup · local validation · MIT-licensed code · fictional sample dataARTICLE 50 OPERATING LAYER
A record is useful only when its claims can be checked.
The Commission separates provider-side machine-readable marking and detection from deployer-side visible disclosure. This starter does not collapse those layers: each control records its layer, method, owner, status, evidence references, and test time separately.
The initial Code signatory-list cut-off is 27 July 2026 at 18:00 CEST; relevant Article 50 obligations apply from 2 August 2026. Later Code signing remains possible.
ONE SMALL, AUDITABLE STARTER
Readable by people.
Checkable by CI.
Provider, deployer, both, or unresolved
Layer, status, owner role, method, and last test
Typed references, capture times, and optional hashes
Next review, accountable role, and decision notes
VALIDATOR COVERAGE
Six checks that fail loudly.
The CLI exits with code 1 on invalid JSON, missing files, broken evidence references, or structural errors, so it can sit in a local review or CI step.
- ✓Required fields and supported enum values
- ✓Unique control and evidence identifiers
- ✓Evidence references that resolve to real entries
- ✓UTC timestamps and lowercase SHA-256 shapes
- ✓Evidence attached to implemented and tested controls
- ✓A test timestamp for every tested control
HONEST TECHNICAL BOUNDARY
The manifest records controls. It does not perform them.
A valid file does not add provenance metadata, watermarks, detectors, or visible labels to content. It does not establish effectiveness, interoperability, robustness, reliability, Code adherence, certification, or legal compliance.
Do not store personal data, credentials, signing keys, confidential prompts, model weights, or sensitive incident material. Use controlled internal references and have accountable reviewers examine the actual system.
FROM SAMPLE TO OPERATING SYSTEM
Need editable workflows around the record?
The free code starter is the implementation sample. The existing price hypotheses remain $39 for the AI Transparency Launch Kit and $69 for the Sprint Bundle; live checkout is still pending storefront approval.