AI Transparency Labels

Every report, advisory, and post we publish can carry an AI Transparency Label — a small structured disclosure that documents how AI tools shaped the artifact. Labels are issued at the moment of publication and pinned at a stable URL on this site, so any reader can audit our disclosure independently.

What a label discloses

  • Involvement level — where on a five-rung ladder the artifact sits, from human-only to AI-generated, unreviewed.
  • What AI touched — research, structure, prose, figures, code, translation.
  • Models used — explicit model names and versions.
  • Human review — none, spot-check, full review, or SME sign-off.
  • Sources — human SME, public web, internal research, client data.
  • Issuer, date, optional artifact hash — to anchor the label to one specific version of the artifact.

Why we publish this

Disclosure is cheap to claim and easy to dilute. Publishing a fixed rubric, a stable URL per artifact, and a public verification page makes the claim auditable. We want our readers to be able to challenge it.

Verify a label

Open cics.ch/ai-label/<id> for any label ID printed on or linked from an artifact. The page renders the full disclosure for that specific artifact.

Generate a new label

Open the label generator →

Vector legend

Each label carries a compact canonical string (AITL vector) that encodes every dimension deterministically. Read the vector legend →

Verify a signed label

How to verify with GPG →