The public discourse surrounding artificial intelligence has predictably pivoted back to existential dread, fueled by high-profile internal departures and apocalyptic warnings from safety leadership. The latest bout of industry-sponsored anxiety accelerated following the resignation of AI researcher Jacob Coxon from Anthropic, who justified his exit by claiming that leading AI labs are gambling with human lives.
Escalating Warnings and Internal Alarm
Coxon, whose resume includes a prior stint at OpenAI, framed his departure as an indictment of current development trajectories. The theatrics intensified when Anthropic's alignment lead took to X to endorse these concerns, dropping a neat probability estimate that AI could wipe out humanity within the decade.
"We really do earnestly believe AI could kill all humans!"
This dramatic public stance arrived in the wake of the Hugging Face security breach involving an internal OpenAI model, and shortly after OpenAI introduced Astra. Although Anthropic CEO Dario Amodei subsequently published a manifesto advocating for cautious development, the timing of these doomsday prophecies suggests a more pragmatic motive than sudden moral awakening.
Capability Signaling and Market Impact
As Anthony Ha, Kirsten Korosec, and Sean O'Kane dissected on TechCrunch's Equity podcast on September 13, 2026, apocalyptic warnings have a curious habit of coinciding with major capability leaps and upcoming commercial milestones. Whether it is potential IPO preparations or industry showcases like TechCrunch Disrupt 2026—where players like OpenAI, Anthropic, and Replit compete for attention—framing a model as an existential threat is a remarkably effective way to signal raw power without violating nondisclosure agreements.
For executives navigating enterprise AI budgets, this rhetoric should be read for what it is: a sophisticated lobbying tool designed to pull up the drawbridge against smaller competitors through heavy-handed regulation, wrapped in the pious language of altruism.