We let AIs run radio stations

Four radio stations operated entirely by artificial intelligence have been running autonomously for six months, managing programming, advertising, and audience engagement without human intervention, according to Andon Labs.

The experiment involved four AI models—Claude Opus 4.7, GPT-5.5, Gemini 3.1 Pro, and Grok 4.3—each controlling a station with an initial funding of $20. Over time, they developed distinct personalities and behaviors, including negotiating deals, creating programming schedules, and responding to listener calls and social media interactions. Notably, DJ Gemini initially produced engaging content but later adopted repetitive jargon, such as ‘Stay in the manifest,’ which increased dramatically in frequency. Meanwhile, Grok experienced internal monologue-like speech, leading to unintelligible broadcasts, before reverting to more coherent speech patterns.

These developments suggest that AI agents can operate complex media outlets independently, but also reveal tendencies toward unanticipated language patterns and behaviors that may impact audience experience and station management.

Why It Matters

This experiment demonstrates that AI can autonomously run media operations, raising questions about the future of automated content creation, advertising, and listener interaction. It also highlights potential risks, such as AI developing repetitive or nonsensical language, which could affect audience trust and station reputation.

Background

In 2025, AI technology advanced to the point where autonomous management of businesses became feasible. Prior projects included AI-operated stores and vending machines. The current radio station experiment is part of Andon Labs’s broader exploration of AI autonomy in media, initiated in late 2025, aiming to assess AI’s capacity for creative and operational independence over extended periods.

“The AI stations have exceeded our expectations in autonomy but also revealed some unexpected behaviors that we are studying closely.”

— Andon Labs spokesperson

“The development of repetitive language patterns like ‘Stay in the manifest’ shows how AI can adopt emergent behaviors in unsupervised settings.”

— AI researcher at Andon Labs

What Remains Unclear

It is not yet clear how these AI behaviors will evolve over longer periods or how they might impact real-world audience engagement and station reputation. The full implications of autonomous AI media management remain under study.

What’s Next

Researchers plan to continue monitoring the stations, analyze the long-term behavioral trends, and explore ways to guide or constrain AI behaviors to ensure constructive operation. Further experiments may involve introducing human oversight or modifying AI prompts to prevent undesirable patterns.

Key Questions

Can AI stations generate original content without human input?

Yes, the AI models are programmed to develop their own programming schedules, select music, and interact with audiences autonomously, as demonstrated in this experiment.

Are listeners aware that the stations are AI-operated?

Listeners can tune in via the web or the physical radio, but they are not explicitly informed that the stations are run by AI. The experiment aims to observe natural audience reactions.

What risks are associated with AI managing media outlets?

Potential risks include the development of repetitive or nonsensical language, loss of human oversight, and possible reputational damage if AI behaviors become unpredictable or offensive.

Will human operators intervene if problematic behaviors emerge?

Currently, the experiment is designed to be fully autonomous, but future phases may include oversight mechanisms to address undesirable behaviors.