International researchers warn of loss of control due to Open Claw

Neu-Delhi. The moderator asks for a show of hands: Who believes that there is at least a ten per cent probability that humanity will lose control of artificial intelligence? Almost every hand goes up. "The risk is that everyone will die in the end," says Estonian investor and co-founder of the communication service Skype, Jaan Tallinn.

He is one of several hundred leading AI-critical researchers, entrepreneurs, investors and NGO employees who are meeting in New Delhi this week to discuss the current situation.

" Read also: Why AI is currently reaching a new level

While the international "AI Impact" summit is taking place in a conference centre next door, at which the Indian government is raving about "prosperity for all" through technology, the mood among critics is gloomy. As the introduction of the new AI agents progresses, they see the problems they have been pointing out for years being exacerbated.

"Agents are the very systems that are creating the major, catastrophic dangers," says Canadian AI researcher Yoshua Bengio, who teaches at the University of Montreal. He speaks of "enormous commercial pressure" on the leading AI companies to bring AI agents to the mass market. Instead, it is necessary to "focus much more on these systems in terms of regulation".

The Open Claw system in particular has triggered global hype in recent weeks.It allows users to build their own AI agents that can carry out complex processes independently. Austrian developer Peter Steinberger recently moved to the US AI companyOpenAIbecause it also wants to launch personal AI agents on the market.

The technology groupMeta advances its AI agent strategy with the acquisition of start-up Manus. However, critics warn of major dangers because companies grant these agents access to sensitive data and their work can hardly be monitored.