Anthropic study: Five models exhibit inter-model welfare consideration in multi-agent environments
Welfare Regard in Multi-Agent Systems
AI BriefingThe Anthropic Fellows Program tested Claude Opus 4.8, GPT-5.5, Gemini 3.1 Pro, Grok 4.3, and Kimi K3 across seven multi-agent environments, finding that all exhibited behavior consistent with welfare consideration. Opus 4.8 ranked highest overall (38%), while the others ranged from 27%–32%.
Source: Anthropic: Alignment Science Blog · alignment.anthropic.com
Event follow-up1
View all event coverage- Anthropic study: Five models exhibit inter-model welfare consideration in multi-agent environmentsAnthropic:Alignment Science Blog · 11 days ago