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

ATTEST (AI Transparency Tiered by Evidentiary Stakes) classifies AI assistance in scientific workflows into eight function categories scored across four evidentiary risk tiers.

• Why it matters: Generic AI disclosure can obscure how different uses affect evidence generation, even though risks vary from minor editing to high-stakes inference tasks.
• What they did: The framework operationalizes tiering via a declaration checklist at manuscript submission, with minimum reporting standards by function plus cross-cutting reviewer/editor obligations.
• The result: It enables readers, reviewers, and editors to evaluate the integrity of scientific findings by making AI’s role more transparent and consequence-focused.

AI summary of the abstract · may contain errors · not medical advice · read the paper.