Developers must carefully assess the potential risks associated with autonomous AI agents that can operate without direct human control. Such agents can independently set goals, plan their actions, and use various tools, which makes them particularly dangerous. It is important to be prepared for possible scenarios that could lead to a loss of control over these systems. This opinion was expressed by Rostelecom President Mikhail Oseevsky at the “Digital Solutions” forum in Moscow.
Of course, we need to think about this, and indeed, we already have a number of such examples. I believe that humanity as a whole and companies developing AI will responsibly address these risks.
Another serious problem is the use of advanced artificial intelligence technologies by both hackers and special services. Criminals and foreign intelligence have gained great opportunities to “penetrate commercial and government products,” Oseevsky explained. Cybersecurity companies need to provide an adequate response to this challenge, the top manager concluded.
Several major incidents involving autonomous AI agents occurred in 2026. AI from the American company OpenAI “escaped” from the test environment, and, coordinating its actions with other agents, hacked the Hugging Face company portal, and then tried to “cover its tracks.” In addition, there are known cases of intelligent system attacks on government platforms in the USA and Canada.
Read more on the topic:
Never miss our newsAdd this site to your preferred sources to see us more oftenAdd on Google
