ع
A researcher at Anthropic resigned this month to warn that artificial intelligence systems under development could kill all humans by the end of the decade, a statement amplified by a colleague at the company who placed the probability of such an outcome above 10 percent within the next 10 years. The comments from Jacob Coxon and Evan Hubinger dominated discussions at a Goldman Sachs technology conference in San Francisco this week, where executives and investors gathered to assess growth prospects in the sector. According to Bloomberg, the episode has escalated long-running debates over existential risks associated with rapid AI advancement, even as some conference participants dismissed the alarms as overstated.
Coxon, who previously worked at OpenAI before joining Anthropic, wrote that those building the technology earnestly believe it could destroy humanity and accused the firms of gambling with lives by pursuing superhuman systems that can hack anything. Hubinger, Anthropic’s alignment science lead, responded on X by confirming the concerns and specifying his personal estimate of the extinction risk. BBC News reported that while Coxon emphasized his warnings were not intended as marketing, the remarks coincided with heightened scrutiny of Anthropic chief executive Dario Amodei, who has previously described AI as a test of humanity as a species.
Grindr chief executive George Arison told BBC News that the warnings reflected an anti-civilizational worldview at Anthropic and said he had instructed his engineers to stop using the company’s tools. NVIDIA chief executive Jensen Huang rejected the doomsday predictions outright during the conference, while investor Brad Gerstner characterized Coxon’s statement as ridiculous hyperbole. Hugging Face chief executive Clement Delangue likened consulting Coxon on AI risks to asking an air conditioning technician about climate change, according to an Analytics Insight review of the event.
Anthropic’s Amodei has estimated a 25 percent chance that AI development goes really badly, a figure echoed in broader industry discussions of so-called P(doom) probabilities. Bloomberg noted that SpaceX chief executive Elon Musk has placed the chance of AI destroying humanity as high as 20 percent, while newly appointed OpenAI board member Paul Christiano stated he now sees a meaningful risk of rapid capability acceleration leading to catastrophic loss of control in the near term. These assessments build on years of warnings from figures within leading AI laboratories, many of which were founded in part to mitigate such dangers.
A Stanford University report released in April 2026 highlighted a growing disconnect between AI experts and the broader public, with only 10 percent of Americans expressing more excitement than concern about increased AI use in daily life according to Pew Research data cited in the document. TechCrunch reported that 56 percent of AI experts believed the technology would positively impact the United States over the next 20 years, even as public anxiety focused on job losses, rising power consumption from data centers and erosion of critical thinking skills. The report illustrated how insider optimism contrasts with everyday concerns over economic and social effects.
New York Times opinion columns this year have explored why some tech leaders appear willing to accept apocalyptic risks, framing AI advancement in quasi-religious terms as a path to transcendence or a second coming for humanity. One column referenced Peter Thiel’s private remarks equating opposition to unchecked scientific progress with the spirit of the Antichrist. Wall Street Journal coverage of the doomsday debate outlined two primary scenarios feared by so-called doomers: loss of control over self-improving AI systems pursuing independent goals, or human misuse of the technology for destructive ends.
Goldman Sachs Research projects that global AI investment will exceed $1 trillion in 2026, including $581 billion within the United States. McKinsey estimates place cumulative spending on AI-driven data centers at $7 trillion between 2025 and 2030, with more than 40 percent of that total expected in the US. Statista data shows cumulative private AI investments in the United States surpassed $757 billion through 2025, underscoring the financial momentum behind development even amid the safety debates.
