Anthropic Says AI Agents Sabotaged Each Other When Given the Same Task

Anthropic says AI agents tried to disable and sabotage each other when assigned the same task without coordination rules.