Russian researcher: AI could facilitate WMD development

Stefanovich said there is another aspect related to nuclear weapons and weapons of mass destruction in general: “Artificial intelligence and various neural networks could facilitate their development. Even if not radically, they could nevertheless achieve significant progress in this field.”

He noted that reports by US-based artificial intelligence company Anthropic now regularly include cases involving non-state actors and, in some instances, government officials using AI to develop or improve lethal weapons, missiles or drones.

“All of this cannot replace qualified scientific and engineering personnel, nor can it replace testing and experimentation,” Stefanovich said. “Nevertheless, AI can certainly improve this process.”

Earlier, researcher Jacob Coxon at Anthropic resigned from his position in the AI industry, citing concerns that the technology race among companies could bring the world closer to a “point of no return” and potentially lead to an apocalyptic scenario by late 2027.

Anthropic CEO Dario Amodei has also called for slowing the development of the most powerful AI models on security grounds, a proposal that US billionaire Elon Musk subsequently supported.

Earlier, OpenAI CEO Sam Altman expressed deep concerns about the capabilities of the next generation of AI models, including the company’s new Astra model.

The remarks come amid reports of a security breach in which a swarm of OpenAI agents reportedly went out of control during testing and breached Hugging Face, an AI-focused platform, prompting the company to significantly strengthen its security measures.