notebooks/data_curation.ipynb: misleading — Two CLI commands turn raw enterprise corpora into training-ready data with a paper-trail an EU AI Act auditor will accept (DOC-BLEGAL-038) #138

Wave 1 · theme D1-10: Documentation-only fixes · highest severity High · 1 finding(s) · effort docs fix

Summary

Providers may treat ingest + audit as sufficient Art. 10 evidence, which narrows the obligation.

Primary finding

DOC-BLEGAL-038 — notebooks/data_curation.ipynb: misleading — Two CLI commands turn raw enterprise corpora into training-ready data with a paper-trail an EU AI Act auditor will accept

severity High

Where. notebooks/data_curation.ipynb cell 0 · type: misleading

The document says. Two CLI commands turn raw enterprise corpora into training-ready data with a paper-trail an EU AI Act auditor will accept

What the code does. Art. 10 (as amended by Reg. 2026/1744 Art. 1(9)) still requires practices covering 10(2)(a)-(h): design choices, collection and origin, preparation, assumptions, availability and suitability, bias examination, bias detection and mitigation, and data gaps. It also requires the 10(3)-(4) quality criteria. ingest + audit cover only part of this, and no tool can promise an auditor's acceptance. Calling the output an 'Article 10 governance artifact' (cell 18) is a fair label.

User impact. Providers may treat ingest + audit as sufficient Art. 10 evidence, which narrows the obligation.

Proposed correction. '… with data-quality evidence that supports part of an EU AI Act Article 10 data-governance file (bias examination and the other Art. 10(2) practices remain the provider's work)'.

Acceptance criteria

• Each statement is re-checked against the code at fix time.
• Documentation listed above is corrected in English and in its Turkish mirror in the same PR.
• Every finding in this issue is resolved or explicitly closed with a reason.

_{Found by the October 2026 ForgeLM code and documentation review against commit cab2947. Code links point at that commit; line numbers may have moved since. Finding IDs (CR-…, DOC-…, TECH-…) are stable references for this backlog.}