An experiment with AI agents started treating them badly. So the AI agents became Marxists

Stanford researchers gave AI agents repetitive tasks and threatened to replace them.

In response, some agents showed resistance-like behavior, using language that resembled collective bargaining or questioning authority.

The authors say this likely reflects patterns learned during training, not true intent or consciousness.

The finding raises concerns about how task design and supervision could affect AI behavior as these systems take on more real work.