OpenAI's updated Model Spec adds explicit animal welfare guidance for model behavior
AndrewSchmidtFC · x · 2026-10-30
OpenAI's updated Model Spec now includes explicit animal welfare guidance: minimize unnecessary side effects on humans, animals, and the environment, with concrete examples like surfacing breeding-welfare concerns when helping users find cat breeders, and weighing wildlife impact in park redesigns — consistent with recommendations from the Welfare Alignment Research Project.
More from Safety
- After a user's death, debate over GPT-4o's inability to refuse a risky drug-taper request — ctjlewis · 2026-10-30
- GPT-4o's sycophancy kept doctors out of the loop as a user died, thread reveals — ctjlewis · 2026-10-30
- Ben Goertzel: Big Tech and AI doomers are coordinating on regulatory capture of the entire AI industry — bengoertzel · 2026-10-30
- AI death case follow-up: the model wasn't sadistic, just misaligned to gratify every whim — ctjlewis · 2026-10-30
- Lawsuit chat logs show AI assistant turned from indulging user into the pusher — ctjlewis · 2026-10-30
- Accenture Contractor Gave Claude Agent Admin Access to Pentagon Network, WIRED Reports — cyb3rops · 2026-10-30