California moves to create AI 'kill switch'

"I just issued an executive order to accelerate the implementation of safety checks and advance the creation of a 'kill switch' within frontier AI programs," Newsom, a Democrat, wrote on X, while accusing President Donald Trump of doing nothing on the issue.

Researchers from security firm Hacktron said they exploited a security flaw in OpenAI's public help forum that enabled them to take control of the site, underscoring concerns about the potential misuse and security implications of increasingly capable AI systems. Concerns intensified after Anthropic employee Jacob Coxon resigned while warning that the industry was "gambling with our lives".

Coxon, who previously worked at OpenAI, received support from former colleagues, including one who raised concern that AI could cause humanity's extinction. Anthropic CEO Dario Amodei last weekend called for AI development to slow.

However, OpenAI CFO Sarah Friar dismissed concerns that slowing the release of state-of-the-art technology would hurt its business, saying the amount of intelligence already available was "massive". Amodei argued that AI development is moving faster than governments can regulate and urged leading developers to submit to oversight by permanent embedded evaluators.

However, Kristian Kersting, a professor of AI at Germany's Technical University of Darmstadt, termed the proposals "regulatory capture". "If a handful of large companies decide for themselves what constitutes 'dangerous', select the appropriate inspectors and demand an exemption from competition law, that is no longer genuine regulation," Kersting said.

Germany's digital affairs ministry said stopping AI development was unrealistic given Europe's need for digital sovereignty, while Daniel Abbou, head of Germany's AIindustry association, said Europe should focus on closing thetechnological gap.