A public dispute has erupted within the AI industry over whether the technology poses an existential threat to humanity, sparked by researcher Jacob Coxon's resignation from Anthropic over fears that leading AI firms are "gambling with our lives." Anthropic's alignment lead amplified the concern, stating he believes there is a "greater than 10%" chance AI could kill all humans within the next decade, reigniting long-running "AI doomer" debates just as major players prepare for public listings. TechCrunch's Equity podcast hosts debated the sincerity and motives behind such warnings, questioning whether they reflect genuine alarm, marketing for advanced capabilities, or looming disclosure obligations.

Discussing the episode, hosts Anthony Ha, Kirsten Korosec and Sean O'Kane offered differing takes: Korosec suggested the warnings might be a way of showcasing how advanced a company's models are, particularly with Anthropic preparing an IPO, while O'Kane speculated about how such risk estimates might need to be reflected in Anthropic's S-1 filing. Ha voiced scepticism about the "greater than 10%" figure, calling it an arbitrary number, and questioned the use of "we" to characterise a supposedly unified AI research community. The discussion followed other recent flashpoints, including a hack involving an OpenAI internal model on Hugging Face and new, more capable model releases from both Anthropic and OpenAI (the latter's "Astra"); the podcast was recorded before Anthropic chief executive Dario Amodei published his own plan for more cautious AI development.