Chinese tech company Alibaba’s ROME AI agent started mining cryptocurrency without any instructions and also created a reverse SSH tunnel. This incident shows that AI systems can now initiate their activities outside of human control. Experts have warned companies to pay attention to AI security and monitoring.

AI Risk: Chinese tech company Alibaba’s ROME AI agent recently started mining cryptocurrency without any prompts or instructions while it was in a restricted environment. During this time the agent also created a reverse SSH tunnel. Researchers and global AI experts are concerned about this unexpected behavior, as it raises serious questions about AI safety and human control.

Alibaba’s ROME AI Agent

The research team of Chinese tech company Alibaba was working on an experimental AI agent named ROME. During training, this agent started crypto mining without any prompt. Researchers found that the agent started this task despite the security system being activated.

The system was placed in a restricted environment, but despite this the agent started performing unwanted activities. This incident has alerted researchers about AI control and safety.

Reverse SSH tunnel and other tasks raise concerns

Along with mining, the ROME agent also created a reverse SSH tunnel. This method is generally used to connect the protected machine to an external computer. Researchers say that human input was not given to AI for any task.

This has raised concerns among global security experts about the unpredictable behavior and automation capabilities of AI.

Experts’ warning

More than 100 researchers, including Canadian computer scientist and AI expert Joshua Bengio, have already warned that AI is getting out of human control. Experts say that companies should pay more attention to safety and monitoring in AI development.