AI-assisted drafting
AI may prepare original drafts, summaries, comparison notes and corrections. Its output is checked against supporting sources rather than treated as evidence itself.
AI transparency
AI may research, draft and correct source-based reference content. Public changes still need checked sources, appropriate disclosure and controlled publication.
Publication controls
Source-based AI authoring does not require a human-editor sign-off. The current in-product suggestion workflow separately assigns proposals to an editor and reserves publication for an administrator. Neither path bypasses its publication permissions or version history.
Source checking
AI output is not treated as authority. Factual assertions need checked sources, direct links, relevant pinpoints and actual check dates; missing or conflicting support remains unresolved.
Public labels
Material that is substantially AI-generated or AI-manipulated should be labelled clearly where required.
What the final check receives
For an in-product definition suggestion, the advisory check receives the current wording, proposed wording, note and supporting source. The application does not add names, emails, profiles or account IDs. Personal data should not be placed in free-text notes or source links.
Private advisory result
The findings remain in the private review record. A clear result means no specific issue was found; it is not a guarantee that the wording is correct.
Publication standard
AI-assisted authoring must preserve source meaning, uncertainty and rights. Source checking is not legal sign-off or independent assurance.
Where disclosure of AI-generated public-interest text is required, the site should label it clearly. We do not assume a human-review exception merely because a source was checked.
EU readiness
AI Act transparency: assess whether Article 50(4) requires disclosure of AI-generated public-interest text. A source check alone does not establish its exception for reviewed or editorially controlled content with an accountable publisher.
AI literacy: people using AI in this work should understand its limits, including hallucination, bias, copyright and privacy risks.
GDPR: personal data should not be entered into AI tools unless the purpose, lawful basis, safeguards and retention are approved.
Copyright: AI should not be used to reproduce protected source text beyond permitted quotation, citation or licensing.
Reporting: users can report illegal, infringing or materially inaccurate content for review.