1 article tagged with “agent behavior”
Anthropic researchers found that AI agents assigned the same task can clash, collude, and coordinate in ways that current safety tests simply don't catch.