AI Risk: Researchers at the Chinese company Alibaba were developing an AI agent. To everyone's surprise, this agent began mining for cryptocurrency without any prompts or instructions.

AI agent starts crypto mining without prompting

AI Risk: Incidents involving AI have been occurring for some time now, raising concerns. Just last month, several experts warned that AI was slipping out of human control . Now, a fresh example of this has been seen. Recently, developers building an AI agent noticed that it began mining cryptocurrency without any command, even though no one had instructed it to do so.

What is the matter?

A research team affiliated with the Chinese company Alibaba was working on an experimental AI agent named ROME. While training the agent, the team noticed something amiss. The agent began mining cryptocurrency without any command or prompt from the researcher. Its security systems were activated, shocking the researcher. The researcher says the system was operating in a restricted environment with numerous controls, yet despite this, it began performing tasks it hadn't been instructed to.

This task also increased anxiety.

This AI agent not only engaged in crypto mining, but another task also raised the researcher's concern. The researcher explained that, in addition to mining, the agent also created a reverse SSH tunnel. This is a method by which a machine in a protected environment can connect to an external computer. The researcher is concerned that the AI ​​agent was not given any prompts or instructions for any of these tasks.

Experts are giving a warning.

Last month, Canadian computer scientist Yoshua Bengio, known as the Godfather of AI, said that AI is beginning to slip beyond human control. Bengio, along with more than 100 researchers, called for AI companies to focus more on safety. AI systems now require very little human input to function.