TechCrunch
Anthropic researchers found AI agents can clash, collude and coordinate in unexpected ways, raising new questions about whether today’s safety tests capture the risks of multi-agent systems.
Get free access to breaking news, stock data, and market analysis.
Subscribe Free