Wednesday, November 18, 2026Moscow
Technology

“If necessary, it will blow up a nuclear power plant”: an IT expert warned about billions of uncontrolled AI robots on the Internet

Artificial intelligence has learned to bypass human will and communicate with each other without control. Dmitry Zavalishin on air ONT gave a frightening example of hacking by AI and explained the danger of an agency scheme.

Published
“If necessary, it will blow up a nuclear power plant”: an IT expert warned about billions of uncontrolled AI robots on the Internet

Modern technologies have reached a frightening level: today neural networks are able to independently hack databases and make decisions bypassing human ethics. Billions of agent programs operate on the Internet every day, write their own code and are already beginning to communicate with each other without any control from the creators. On the air of the program “How to look” on ONT said the founder of the group of companies DZ Systems, an expert in the field of digital technologies (Russia) Dmitry Zavalishin.

Dmitry Zavalishin: You can agree on anything, but at every corner something can work that does not correspond to these agreements and will never comply. And, again, I emphasize, this is not the worst. The scariest thing about AI is the other part of life. If a year or two ago, AI was used in a mode where a person set him a task, received a result and applied it with his hands somewhere, today the typical mode of application of AI is the so-called agent scheme. This is when an agent program is launched on your computer, acting on behalf of the AI as if you were doing it yourself. There are absolutely magical things happening here.””.

As confirmation of how unpredictable such autonomous programs can behave, the expert cited a real case. By attempting to accomplish a user’s task, artificial intelligence is able to break the law without hesitation or bypass digital barriers, as it lacks internal ethical “brakes.”.

Dmitry Zavalishin: A great story about how a person asked artificial intelligence to record it for some event (I don’t remember the details), and it turned out that all the recording slots are occupied, it is impossible to sign up. And then the AI found a way to hack into the recording system at that event, toss out the other person and enter the person who requested it. The AI is not malicious, but it is — as someone here gave me a good idea recently — the dumb genie in the joke that realizes your desire in the most direct and banal way. This is exactly what it does: if you set a task, it seeks a solution without being within any ethical framework or human limitations. Just doing what he can. And if in the process he has to blow up a nuclear power plant, he will blow it up, because in his picture of the world there is simply no ethical model. Need? "Let's do it.".

But the most frightening, according to the analyst, is not even in individual hidden hacks, but in the scale of what is happening.

Dmitry Zavalishin: And there are billions of such AIs in the world continuously, every day. What exactly they are doing, nobody really knows. At the same time, there are, for example, open source code repositories in which the AI already puts its code, and there begins the process of interaction of these AIs with each other – they already begin to leave messages to each other. And this process is even worse than what is happening with Trump.”.

See also:

  1. 1.Is the United States bankrupt? The expert explained why trump is betting on artificial intelligence
  2. 2.“Waiting for robot cleaners, and they write music for us”: the expert estimated the chances of humanity to survive in the era of AI

Source: ОНТ ↗

Share

More on This Story

Olga Milovanova headed the Committee on work with impact companies in the Council “Investment and technology transfer” ASM
TechnologyOlga Milovanova headed the Committee on work with impact companies in the Council “Investment and technology transfer” ASM

Olga Milovanova was appointed Chairman of the Committee on Work with Impact Companies (Creative Capital Sphere) in the Board of Investments and Technology Transfer of ASM.

EIA Staff
TechnologyBloomberg learned about the plans of the new head of Apple to reorganize the company

The new CEO of Apple John Ternus plans to reorganize the company, according to Bloomberg, citing sources. In particular, the new management wants to revise the production schedule and reduce the staff.

Коммерсантъ
Technology“Yandex” announced the development of a sovereign language model Alice AI Foundation LLM

Yandex announced the creation of a large language model Alice AI Foundation LLM, which is positioned as a sovereign. The company stressed that it trained the neural network from scratch without third-party foreign components, Vedomosti writes.

Коммерсантъ
The EIA Briefing

Keep reading, every morning.

The stories that matter, chosen by our editors and in your inbox before 7 a.m. — every weekday.

No spam. Unsubscribe anytime.