Article 50 Transparency Explorer

Historical coded values for twelve consumer AI products, with declared collection on 5 and 6 August 2026. Click a cell to inspect its original coding and source note. These stored values are preserved by the version 1.1 correction; they are not a current survey, an accepted rescore or a compliance assessment. Capture dates and account states have not been independently authenticated.

AI assistance: original work with Anthropic; retrospective correction with OpenAI GPT-6; separate bounded Gemini advice. These providers have products in the scored set. The author retains analytical responsibility. Review is not exhaustive.

Read the correction before using the scores.

Source support remains unresolved for Copilot 2.3 and FLUX, Gemini and Synthesia 2.4. Disclosure-name interpretation and capture/placement also retain explicit limits. A positive stored value is not proof of current implementation or technical performance. Correction note.

Paper (DOI) Code & data NM AI Research

5/12historical yes values for interaction disclosure (1.1)
2/12historical yes values for coverage (2.4); support unresolved
0/12historical yes values for accessibility documentation (5.2)
Proposed repeat3 to 9 December 2026; separate approval required
✓ historical yes value (not newly validated source support) ✗ historical no value ○ historical no artefact found (a search outcome)

The 50(7) column is the Code of Practice signatory status. It is reported beside the table and is never added to the ten-item total. “no” is drawn neutral, not red: this measures public evidence, not compliance.

Later comparison is a separate study.

The proposed collection window is 3 to 9 December 2026 UTC. Execution requires separate author approval, matched product identities and account states, and suitable validation for the new data. A missed window requires a dated amendment. The original search budget and baseline states cannot all be reconstructed. No second survey has been run for this correction. Repeat protocol.