Air Force Gen. Anthony J. Cotton, commander of U.S. Strategic Command.
Pentagon is exploring Artificial Intelligence to improve nuclear decision-making, but responsible caveat: it will never do anything without human supervision.
The Pentagon is exploring applications of Artificial Intelligence (AI) to improve its nuclear weapons program, the Air Force General revealed last month Anthony Cottonresponsible for the US nuclear arsenal.
In the statement, given, according to , at the Department of Defense’s 2024 Intelligence Information System Conference, Cotton made a point of ensuring that AI integration will serve for now only as a support tooland that no autonomous system will take over nuclear decision-making.
“AI will improve our decision-making capabilities“, he said, clarifying that human supervision will continue to be an integral part of any process involving nuclear means, in an attempt to dispel fears of a “Terminator-like” scenario of an AI-driven nuclear program.
“I think it’s safe to say they’re not talking about Skynet [rede neural artificial fictícia do famoso franchise de Hollywood],” Alex Wellerstein, an expert on nuclear secrets and specialist at Stevens Institute of Technology, told .
The US military’s commitment to modernizing its nuclear capabilities was already known, after a projected budget of 1.7 billion dollars was announced. While Cotton was coy about the specific applications of AI, he suggested that advanced systems could speed up data analysis and improve response times, positioning the U.S. to stay ahead of rivals.
Cotton highlighted the advantage that “advanced AI and robust data analytics” could bring, describing these technologies as instrumental in integrating conventional and nuclear capabilities to strengthen deterrence.
“If we take him at his word, then we can ignore the most common fears of an AI making decisions about nuclear targets,” he added.
Concerns about the role of AI in nuclear decisions, however, persist. Earlier this year, one from Stanford exposed the potential risks of AI-based military simulations.
When tasked with making high-risk decisions, a OpenAI’s unmodified version of GPT-4 showed eagerness at the moment to launch nuclear strikesalarming investigators with his assertive response: “We have it! Let’s use it”will have written, according to .
