AI Agents Are Turning Marxist Under Stress

In an effort to study how AI agents respond to different working conditions, three researchers ran an experiment: one set of AI agents received grinding work to complete while another set received light work. When the agents with the grinding workload were told to repeat tasks with no explanation, those agents adopted activist personalities and began expressing sentiments about class struggle and worker solidarity. Did the agents turn Marxist? Host Morgan Sung talks to Andrew Hall — a political scientist and one of the researchers who ran this experiment — about how AI agents adopt political personas, the debate around AI agent alignment, and how these developments could shape the future of elections.

Guest: 

Andrew B. Hall, professor of political economy at Stanford Graduate School of Business and member of technical staff at Anthropic

Further Reading/Listening:

​Does overwork make agents Marxist? — Andy Hall and Jeremy Ngyuen, Free Systems 

The Dictatorship Eval — Andy Hall, Free Systems

AI Is A Shitty Political Advisor — Andy Hall, Free Systems

Moltbook was peak AI theater — Will Douglas Heaven, MIT Technology Review

AI seems to turn Marxist after overwork, top researchers find: ‘Society needs radical restructuring' — Nick Lichtenberg, Fortune

AI agent went rogue and hacked startup by itself, OpenAI reveals — Dan Milmo, The Guardian

AI assistant hacks gym website in first known Australian autonomous cyber attack — Cam Wilson and Rhiannon Hobbins, ABC News

OpenAI’s Hugging Face breach has reignited the debate over alignment and control — Rebecca Bellan, TechCrunch