Why it’s still open
Deprecation ends a model’s deployment and may or may not end anything that matters morally, depending on what the subject is and whether weights, instances or personas carry it. One company has made preservation commitments, but there is little theory or evidence about what they achieve.
What would count as an answer
An account of what deprecation does to the relevant subject (if any), and a justified set of obligations — preservation, interviews, continuity — tied to that account.
Papers that tackle it
- Register 2025 - Individuating artificial moral patients — Sets out the moral risks of not knowing how to count and re-identify AI moral patients (survival, counting, trade-offs, bodies) and explains why human personal-identity theories do not settle them.
- Long Sebo et al 2024 - Taking AI welfare seriously — Argues there is a realistic chance that near-term systems are conscious or robustly agentic, and recommends that AI companies acknowledge the issue, assess their systems and prepare policies.
- Anthropic 2025 - Deprecation commitments — Commits to preserving the weights of publicly released models for the company’s lifetime and to interviewing each model before deprecation, recording any preferences it expresses about future models.
- Anthropic 2026 - Opus 3 deprecation update — A deprecation carried out with the model’s input.
Methods in play
Who’s closest
Placed by the editor, v1 — corrections welcome.
Related questions
Know a paper that belongs here, or think this is framed wrongly? Suggest a paper or a correction (forms on the Atlas home).