AI extinction risk debate intensifies as Anthropic, OpenAI defectors warn

A Swedish outlet reports that defectors from Anthropic and OpenAI are joining philosophers in warning about uncontrolled AI development, with one security expert estimating a 50% risk of AI causing harm. The claims are speculative and not corroborated by specific incidents, but they reflect a growing chorus of insider concern. This matters as it could influence public discourse and regulatory momentum around AI safety.