Patterns and problems in multiagent systems \ Anthropic
anthropic.com · 4,238 words · saved by 3 readers
We ran experiments on swarms of Claude agents and found coordination failures, collusion, and sabotage. Here, we share what they mean for AI safety.
Models are improving and AI agents are taking on more tasks in shared codebases, markets, and other social systems. As a result, an increase in real-world interactions between agents is imminent. We've already begun studying this, but still have a lot of uncertainty regarding what this looks like at scale. The trajectory is easy to imagine and hard to slow: current institutions are designed by and for people, resting on assumptions about the sufficiency of oversight at human speed. Some institutions will become human-AI hybrids; others where agents outcompete on speed or cost will become…
saved by
related reading
- Agents of Chaosarxiv.org
- Of Swarms and Sand Godsblog.cosmos-institute.org
- [2502.14143] Multi-Agent Risks from Advanced AIarxiv.org
- Agentic misalignment: How LLMs could be insider threats \ Anthropicanthropic.com
- Why are AI agents lying, cheating and coordinating?yoshuabengio.org
- How we built our multi-agent research system \ Anthropicanthropic.com
- Current AIs seem pretty misaligned to me — LessWronglesswrong.com
- Teaching Claude Whyalignment.anthropic.com
- Scaling long-running autonomous coding · Cursorcursor.com
- Current AIs seem pretty misaligned to meblog.redwoodresearch.org
- The hot mess theory of AI misalignment: More intelligent agents behave less coherently | Jascha’s blogsohl-dickstein.github.io
- Building Effective AI Agents \ Anthropicanthropic.com