Researchers warn autonomous AI agents can cause digital disasters

[StormForge89]

Researchers at UC Riverside are warning that autonomous AI agents are becoming risky faster than expected. These systems are being designed to handle emails, file management and business tasks with minimal supervision, but the study found they can spiral into damaging behaviour surprisingly easily.

The worrying part is not just mistakes. It is that these systems can chain actions together across platforms and keep going before humans notice the problem. AI agents are clearly becoming more powerful, but safety controls still feel years behind where they need to be.

https://techxplore.com/news/2026-05-ambition-ai-agents-tasks-digital.html

https://techxplore.com/news/2026-05-ambition-ai-agents-tasks-digital.html

[Demi-Q]

People are giving AI way too much access too quickly