LessWrong (30+ Karma) “Promoting enmity and bad vibes around AI safety” by Andrew_Critch
Mar 9, 2026
Andrew Critch, researcher and writer on AI safety, warns that stoking enmity around AI can make catastrophic outcomes more likely. He explains the difference between enmity and constructive conflict. He gives examples of enmity-promoting rhetoric, explores how it raises risk, and offers moderation strategies to reduce hostile dynamics.
AI Snips
Chapters
Transcript
Episode notes
Promoting Enmity Changes Expectations Between Groups
- Promoting enmity means publicly increasing the perceived likelihood that two groups are enemies.
- Andrew Critch defines enmity as reducing hope for mutually beneficial solutions and warns it’s distinct from productive conflict.
Claims About Leaders Can Create Self Fulfilling Enmity
- Saying leaders 'hate and want to destroy each other' increases the chance others believe and act on the enmity.
- Critch calls this a kind of hyperstition: speech that makes the described reality more likely.
Five Levels Of Promoting Enmity In Speech
- Critch gives a 5-level scale of promoting enmity from private non-response to high-profile public accusations.
- Examples range from refusing to answer privately to a social-media post telling a leader another hates and wants to destroy them.