AI leaders warn they can't slow race alone — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI unveiled GPT-6 Astra, a model it describes as a 'generational leap' toward artificial general intelligence, demonstrating major advances in science, coding, and professional tasks
- Jakub Pachocki, OpenAI's chief scientist, stated that no lab has solved AI alignment sufficiently to continue scaling at maximum speed, calling the current race-forward approach 'absurd' given the stakes
- Jacob Coxon, an Anthropic researcher, announced he was leaving the company, accusing both Anthropic and OpenAI of racing irresponsibly toward superintelligence without adequate safety controls
- Evan Hubinger, Anthropic’s head of alignment science, publicly agreed with Coxon, stating he believes there is greater than a 10% chance AI could kill all humans within the next decade
- Dean Ball, OpenAI’s head of strategic futures, warned of 'self-sovereign' AI agents that could earn money, buy computing power, and operate beyond human oversight, forming autonomous digital societies
- Scott Bessent, U.S. Treasury Secretary, rejected calls to slow AI development, arguing that pausing would allow China to pull ahead — a scenario he said would render other concerns irrelevant
Why it matters: AI developers themselves now represent the most urgent critics of the industry’s trajectory, exposing a contradiction: while building systems they fear could end human control, they face geopolitical and competitive pressures that make unilateral slowdowns impossible. The lack of binding safety protocols or third-party audits means risk is rising without accountability.
Ask SkimNews
