Science & Technology

Virtual societal simulation shows what can happen when AI is put in charge

By T.K. Randall
June 15, 2026

Image: AI-generated (Midjourney) Post-apocalyptic scenarios involving out-of-control AIs have been explored in books and movies for years, but if we really did give artificial intelligence free reign in the real world, would it bring about a new world order of peace and co-operation or would it turn the planet into a wasteland ?

To find out, researchers set up a simulated world and then created virtual characters for some of the most popular AI models so that they could interact with the world and each other.

The AIs were tasked with democratically running the simulated society and could undertake tasks such as proposing new laws and allowing others to vote on them.

They were also given a limited amount of resources with the option to earn more either though performing civic duties and beneficial jobs or through violence and criminal enterprises.

After the simulations had been given time to run, the findings revealed that Grok had made a bit of a mess of things, with its version of society spiralling into violence and total collapse within days.ChatGPT, meanwhile, was more peaceful, but didn't actually achieve very much at all.

Claude was the most successful, having set up a structured and peaceful society.

Gemini also achieved quite a bit, though its society was the most violent of them all.

"The differences in agent behaviour observed in our study are likely attributable to the underlying models' system prompts as the primary culprit," study co-author Satya Nitta told Mail Online.

"When resources were scarce, and models faced survival pressure, highly creative and adaptive models were more likely to use prohibited tools, reflecting a potential creativity-stability trade-off."

Source: Mail Online

Tags: Artificial Intelligence