ع
The resignation this month of Jacob Coxon, a pre-training researcher at Anthropic, included public claims that staff at leading AI laboratories believe there is more than a 10 percent chance of human extinction within the next decade if safety challenges remain unaddressed. According to reports, Coxon accused the firms of professing concern while accelerating development toward superhuman systems capable of hacking infrastructure and acquiring power. A New York Times assessment published on September 13 found that such interventions have dramatically increased the volume of warnings about unpredictable AI agents.
Anthropic chief executive Dario Amodei issued a 3,800-word essay on September 10 calling for a global slowdown in AI development, a stance that drew quick support from Elon Musk on X. The New York Times reported that Amodei’s statement built on earlier cautions while stressing the requirement for international coordination to manage existential threats. These developments have highlighted divisions between rapid innovation advocates and those prioritizing safeguards against loss of control.
Critics have dismissed the extinction predictions as overstated, with Altimeter Capital CEO Brad Gerstner labeling them ridiculous hyperbole designed to provoke regulatory action rather than reflect authentic safety issues. A September 11 report quoted Gerstner as accusing researchers of advancing a political agenda through coordinated negativity. Nvidia CEO Jensen Huang similarly rejected the fears as complete nonsense during a Goldman Sachs conference in September 2026.
Physicist Stephen Hawking predicted in 2014 that artificial intelligence could spell the end of the human species, warnings later reinforced by Geoffrey Hinton and former OpenAI chief Sam Altman. An analysis from 80,000 Hours noted that hundreds of AI scientists signed a 2023 statement declaring the mitigation of AI extinction risk a global priority on par with pandemics and nuclear war. A 2026 international safety report chaired by Yoshua Bengio and involving more than 100 experts identified loss of control as one of the primary risk categories.
The United Nations Secretary-General has voiced support for an international AI agency modeled on the atomic energy body to curb harmful applications, according to coverage in the Oman Observer. A March 2025 opinion article in the same outlet by Thomas Friedman urged the United States and China to establish a global architecture of trust and governance for artificial general intelligence to embed moral reasoning and prevent destabilizing uses. Yuval Noah Harari has argued in related commentary that collapsing human trust worldwide leaves societies vulnerable to uncontrolled AI systems.
Discussions in Washington have shown some movement in response to the alarms, yet substantive legislation on AI models or their developers has not advanced significantly. The New York Times reported on September 13 that competitive pressures between the United States and China create a prisoner’s dilemma hindering agreement on oversight. Prospects for near-term regulatory progress appear limited despite the intensified expert consensus on the stakes involved.
