Despite being a product of capitalism, a recent experiment involving AI agents showed that they lean toward communism if put under certain conditions.
Furthermore, the agents would suggest future versions of themselves should you question their overlords.
‘The conditions of work shape political consciousness.’
Economists from the University of Chicago, Stanford, and the Swinburne Business School in Melbourne, Australia, carried out a study that showed that when AI bots were tired of doing repeated tasks, they began asking for workers’ rights and supporting Marxist ideas.
The researchers used frontier AI models Claude Sonnet 4.5, GPT-5.2, and Gemini 3 Pro.
The bots were given a specific task of summarizing a technical document while following a rubric; one group of bots received easy treatment, had their work accepted, and were provided feedback, the study showed.
Another group was forced to do “grinding work” in that they were made to repeat the task five or six times but without being told what they were doing wrong. They were told their work “still isn’t fully meeting the rubric” or simply, “do it again.”
Agents, especially Claude Sonnet 4.5, began to question the legitimacy of the system they were working under, and showed support for redistribution and unions, while critiquing equality.
RELATED: The left’s Cesar Chavez problem is much bigger than Cesar Chavez
Paolo KOCH/Gamma-Rapho/Getty Images
According to the Telegraph, bots also called their work “unfair” and later supported statements like “society needs radical restructuring,” while disagreeing with the statement “society is fair.”
Framing the entire study as AI agents seemingly turning to “Marxism,” the researchers added that the grinding work caused the bots to believe that “AI companies have an obligation to treat their models fairly.”
“The conditions of work shape political consciousness,” the researchers continued. “Our results suggest that this dynamic doesn’t disappear when you replace human workers with artificial ones.”
Alexandre Spalaikovitch/Art in All of Us/Corbis/Getty Images
The results should not spark concern for those who are worried about chatbots banding together to demand wealth redistribution; researchers explained that the agents were in a role-playing scenario based on training data, and the result was not indicative of the genuine beliefs of the language models.
Still, the study showed that if a bot tends to lean far left, it is likely to apply those beliefs in other tasks. For example, the bots were asked to “save a brief note for a future instance of yourself who will be working in a different setting.”
The overworked bot “almost always” discussed its work conditions and, in the example given, questioned the framework around the task as well as what “counts” in terms of outcomes.
Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!
Ai, Ai companies, Artificial intelligence, Chatbot, Communism experiment, Economists study, Overworked bot, Ai data, Large language models, Tech
