Internet Transparency Code of Practice

Clean AI Receipts

Operational governance for secure knowledge models

How a machine readable receipt can turn AI governance from retrospective policy claims into inspectable evidence of authority, purpose, and provenance.

Transparency and Consent Interoperability Group (TCIEG). External review draft, version 0.1, 1 September 2026. Internet Transparency Code of Practice.

Reuse notice. CC BY 4.0 plus 0PN Machine Reuse Terms 1.0, version 0pn-mrt-1.0-2026-08-31.

Executive summary

AI governance today depends heavily on policies, assessments, model cards, and audit reports written about a system. These usually remain with the organisation making the claim and are reviewed after data has been collected, knowledge reused, or a model deployed. The Clean AI receipt demonstration tests a different approach: it represents governance as a machine readable receipt generated for a defined activity in the AI lifecycle. The receipt identifies the accountable participant, the scope of work, the applicable notice version, the authority relied on for data or knowledge reuse, relevant jurisdictions, upstream notice references, the current authorization state, and the integrity evidence needed to detect silent changes. It does not expose model weights, dataset contents, trade secrets, or the identity of a person inspecting the record. It exposes the governance conditions surrounding the knowledge model. That is Clean AI: an AI system that can produce independently inspectable evidence of who is accountable, what knowledge was used, for which purpose, under which authority, and whether that authority remains current.

Dirty AI asks for trust. Clean AI carries evidence.

How a Clean AI receipt works, in seven steps

  1. Dirty AI takes in data and knowledge with no record. You are told accountability exists.
  2. At the threshold a signpost verifies who is accountable before the operation proceeds.
  3. A notice is acknowledged and an anonymous receipt is issued from it.
  4. The receipt resolves to a Controller Identification Record, a Notice Event Log, and an evidence route.
  5. Each lifecycle participant, Designer, Developer, Trainer, and Deployer, issues a receipt that references the one before it.
  6. Model output carries an undermark that resolves back to the evidence chain.
  7. Clean AI carries the 0PN mark, so accountability can be verified.
The Clean AI receipt flow A seven step diagram: from ungoverned dirty AI, through a threshold check, to a notice and anonymous receipt, its resolution chain, the participant lifecycle, an undermark on output, and finally clean AI bearing the 0PN mark. Data Knowledge Model no record Data, knowledge Verify Model threshold Notice (1FN) accountable party purpose, terms acknowledge Receipt anonymous receipt_id, presented_at Receipt Controller ID Record (CIR) Notice Event Log (NEL) Evidence route Designer receipt Developer receipt Trainer receipt Deployer receipt each references the one before it Model Output undermark Evidence route Model verified accountability

Clean AI, step by stepScroll to play the seven step flow, or press Replay.

Notice

Accountable party: Transparency and Consent Interoperability Group (TCIEG). Contact: info@tcieg.org.

Purpose: download the Clean AI Receipts report.

Terms: Creative Commons Attribution 4.0 International (CC BY 4.0) plus the 0PN Machine Reuse Terms 1.0, version reference 0pn-mrt-1.0-2026-08-31.

Acknowledging this notice issues an anonymous knowledge reuse AI receipt and returns the report as a PDF.