Skip to main content

AI Moral Status Evidence Tracker

AI Alignment Policy Institute · v0 · updated 22 July 2026

Functional evidence only
This tracker records what published evidence shows about functional properties. It takes no position on subjective experience, and it grants no rights and confers no legal status.
Candidate indicator family Behavioral Internal Developmental
Access-consciousness properties
Global workspace theory		 Open Open
Functional valence and affect
Emotion representations
Agency (minimal → rational)*
Three-level taxonomy		 Open
Self-reports and self-advocacy
Apparent self-reports		 Open
Select any graded cell to see what the evidence shows, its key caveat, and the sources behind the grade.
Partialdemonstrated in part, replicated Emergingearly findings, unreplicated or contested Openno substantial evidence either way

* This row grades evidence for the presence of agency-associated properties. Whether agency alone grounds welfare or moral status is a separate, contested question (Long, Sebo et al., 2026); presence is graded here, moral weight is not.

Grades follow the behavioral / internal / developmental evidence taxonomy of Long, Sebo et al. (2026), Studying AI Welfare Empirically. Sources added this revision: Anthropic (2026), Claude Fable 5 & Mythos 5 System Card; Ren et al. (2026), AI Wellbeing (Center for AI Safety, preprint). Phenomenal experience: unresolved in all cells.