ATTEST 2027: A Framework for AI Transparency Tiered by Evidentiary Stakes.

Authors

Wanzhu Tu, Jun Li, Michael Charlton, Naga Chalasani, Laura E Nagy

Published in

Hepatology (Baltimore, Md.). Nov 05, 2026. Epub Nov 05, 2026.

Abstract

Artificial intelligence (AI) is now being used across scientific research workflows, from data extraction and code generation to literature synthesis and manuscript drafting. Scientific journals and professional organizations agree that authors should disclose AI assistance. But disclosure that treats all AI use equally with a single generic declaration reveals little about the nature of AI's influence on evidence generation. Polishing grammar carries a different risk than extracting outcomes from clinical notes, classifying pathology, or making causal inference. We propose ATTEST (AI Transparency Tiered by Evidentiary Stakes), a framework that evaluates AI assistance by its consequences. ATTEST classifies AI use into eight broad categories of functions, each scored at one of four risk tiers (Low, Medium, High, Very High). The framework is operationalized by a declaration checklist that authors complete at submission. Within this tiered system, we propose minimum reporting standards by function, cross-cutting requirements, and obligations for reviewers and editors. The goal is to make AI's role in evidence generation transparent to readers, reviewers, and editors so they can evaluate the integrity of scientific findings.

PMID:
42992833
Bibliographic data and abstract were imported from PubMed on 06 Nov 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.