/Tech2h ago

Google DeepMind's Andreas Kirsch and policy specialist Séb Krier argue multi-agent AI systems introduce emergent misalignment risks over monolithic models

They urge safety research to focus on collective dynamics.

41526682

#271

Original post

Andreas Kirsch 🇺🇦@BlackHC#271inTech

@sebkrier Large scale multi agent collectives should be easier to keep aligned, shouldn't they?

As long as they are not RL'ed together to create their own code language, they should be easier to monitor. Or is there a risk for emergent misalignment?

Séb Krier@sebkrier

How do we go from AGI to Superintelligence? New report discusses four potential pathways: scaling, AI paradigm shifts, recursive improvement, and ASI emerging from large-scale multi- agent collectives. Importantly, it also looks at possible frictions and bottlenecks along these pathways. Instant classic! https://arxiv.org/abs/2606.12683

7:43 AM · Jun 12, 2026 · 434 Views

Sentiment

Sentiment building, check back later.

Cluster Engagement

Posts from X

Most Activity

VIEWS282BOOKMARKS5LIKES8RETWEETS1REPLIES2

Séb Krier@sebkrier

@BlackHC Not personally a fan of the 'aligned/misaligned' taxonomy and framing but yes I think so! There are collective risks too ofc (https://arxiv.org/abs/2512.16856 / https://www.cooperativeai.com/post/new-report-multi-agent-risks-from-advanced-ai).

Andreas Kirsch 🇺🇦@BlackHC

@sebkrier Large scale multi agent collectives should be easier to keep aligned, shouldn't they?

As long as they are not RL'ed together to create their own code language, they should be easier to monitor. Or is there a risk for emergent misalignment?

2h28285