Multi-agent risks go mainstream after UK AISI, OpenAI, Anthropic incidents

sebkrier · x · 2026-09-11

A Johns Hopkins policy researcher recalls advocating multi-agent risks years ago, only to be dismissed with "that sounds like a Seb take" — until the UK AISI, OpenAI, and Anthropic incidents made the topic swarm-worthy.

The post criticizes excessive deference to funders in EA/AI safety circles: "You should convince funders of your position, not the reverse."

Related event: Multi-Agent Risks Research Goes Mainstream, Early Warners Vindicated(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →