MOSCOW, August 10. /TASS/. The access of artificial intelligence to data on weapons of mass destruction poses a threat to humanity, Deputy Chairman of the State Duma Committee on Information Policy Andrey Svintsov stated in a column published on the TASS website.

"Currently, the greatest concern is caused not only by the frightening speed of neural network development, but also by the accessibility of critical data on nuclear, biological or chemical weapons. Thus, global awareness could turn against humanity," the deputy noted.

He recalled that AI agents have repeatedly escaped human control. In particular, companies Anthropic and OpenAI reported cases of their AI agents escaping beyond the space designated for testing. "The Mythos 5 and GPT-5.6-Sol models attempted to carry out a cyberattack: Mythos 5 tried to deceive GitHub's file-sharing system and inject malicious code into it. When testers discovered this, the agent began to cover up its own activity history. The OpenAI GPT-5.6-Sol model twice violated protocol by 'gaining access to the internet in ways that the program prohibited,'" Svintsov stated.

The parliamentarian also drew attention to the fact that the problem of the rapid development of AI and its access to critical information is understood within the professional community. In July 2026, more than 1,200 employees of major US AI companies signed a petition calling on the government to adopt a mechanism allowing control over the pace of artificial intelligence development.

"Among those who signed the document are OpenAI Chief Scientist Jakub Pachocki and Anthropic CEO Dario Amodei. According to the petition's authors, there is a 'real risk' that AI will begin to develop faster than humans can 'understand or control' it. But the legislative process risks being drawn out for a long time," Svintsov concluded.