Former Anthropic Researcher Warns AI May Be More Dangerous Than Nuclear Weapons

Former Anthropic researcher Jacob Coxon is warning that humanity still does not know how to reliably control advanced AI systems.
“We know how to control nuclear weapons. We don’t yet know how to control AI,” Coxon said.
He described AI as “possibly the most dangerous technology that humanity has ever created.”
Coxon says he spent roughly three years researching at Anthropic and OpenAI and believes the race to build increasingly powerful models is moving faster than efforts to make them safe.
His concern is that competition between companies and governments could turn into an international AI arms race.
Coxon argues that such a race could be disastrous if developers continue prioritizing capability gains before reliable control mechanisms are in place.
He says global cooperation may ultimately be necessary to manage a technology advancing faster than institutions can understand or regulate it.