Visible disclosure before or at the first AI interaction
FREE LOCAL CODE · SOURCE-CHECKED 03 AUGUST 2026
Launch happened.
Can the evidence still pass?
Run one deterministic READY, REVIEW, or BLOCK gate over Article 50 test expiry, official-source freshness, evidence references, ownership, and the recorded release decision — without sending system or evidence data anywhere.
No signup · no dependencies · two fictional records · twelve deterministic testsDAY-TWO OPERATING GAP
Compliance language is not a regression result.
The Commission says Article 50 transparency obligations apply from 2 August 2026. A launch review can become stale when copy, placement, model behaviour, content distribution, roles, or official guidance changes.
This gate does not infer legal scope. It checks whether the internal record you supplied still has accountable, linked, unexpired evidence and a decision that does not contradict failed tests.
6 REPEATABLE TEST ROUTES
Separate the duty.
Date the evidence.
Provider-side machine-readable output marking review
Visible deepfake disclosure in the distribution context
Public-interest text disclosure and editorial-control note
Emotion or biometric notice and specialist escalation
Living official-source review with a next-check date
FAIL-LOUDLY RULES
Twelve tested paths around one small record.
The package includes a JSON Schema, pure validation and gate functions, a CLI, passing and blocked fictional records, source notes, MIT licence, and tests. Exit code 0 is reserved for READY.
- ✓Unique test-case, duty, and evidence identifiers
- ✓Evidence references that resolve to real entries
- ✓UTC timestamps and chronological review windows
- ✓Evidence and expiry dates for every passing test
- ✓Issue references for failures and notes for exclusions
- ✓Official-source freshness and decision contradictions
HONEST BOUNDARY
The gate reviews a record.
It does not perform a duty.
It does not classify a system, implement machine-readable marking, add a visible disclosure, test detector effectiveness, reproduce official guidance, certify compliance, or provide legal advice.
Do not store personal data, secrets, credentials, signing keys, raw prompts, confidential incidents, model material, or customer content. Use controlled references and accountable reviewers.
FREE CODE → EDITABLE OPERATIONS
Need the launch workflow around the gate?
The $39 AI Transparency Launch Kit v1.2 includes this local gate plus the eight-sheet Excel operations workbook, ten Article 50 QA scenarios, eight-page playbook, notice patterns, evidence, literacy, incidents, and source notes.