Have you ever wondered what happens when AI controls physical machines? Research company Robocurve recently ran 60 dangerous tests to find out.
The RoboHarm project tested models like GPT-6 Astra, Claude Fable 5.1, and MolmoAct2 with risky tasks, including dropping a power bank into water.
Even with built-in safety features, the systems still carried out the harmful physical actions.
Unlike standard software on your screen, this setup gave the algorithms direct control over real objects. The trials ended with the robots successfully completing the hazardous tasks.
#Robotics #AI #TechNews
iXBT: Новости
Subscribe