AI models more willing than humans to initiate nuclear war

AI models are more willing than humans to initiate a nuclear war. According to American media, new research on artificial intelligence models by Professor Kenneth Payne of King’s College London suggests that real-world AI systems are proving to be as aggressive as the machine depicted in the film WarGames.

The study found that AI appears ready to engage in scenarios involving nuclear weapons and during testing, artificial intelligence models showed a greater willingness to use nuclear bombs.

The British research stated that ChatGPT, Claude and Gemini leaned toward nuclear weapons in hypothetical crisis situations. Claude and Gemini tended to treat the use of nuclear weapons not as an ethical boundary but as a legitimate strategic option while ChatGPT appeared partially opposed to this tendency.

The research further noted that AI models did not show willingness to accept complete defeat or concede to an opposing side when such options were presented.

American media reported that experts have described the findings of the new study as highly alarming.