Former Anthropic researcher resigns, warning AI race 'Could kill us all by the end of the decade'

Marijan Hassan - Tech Journalist

56 minutes ago

2 min read

An explosive debate over existential AI risk erupted on social media following the high-profile resignation of Jacob Coxon, a prominent artificial intelligence pre-training researcher who previously worked at both OpenAI and Anthropic. In a public resignation thread, Coxon warned that leading laboratories are recklessly racing toward self-improving superintelligence without adequate safety controls, asserting that many frontier developers privately believe the technology could cause human extinction before 2030.

Internal Alignment Leads Confirm Existential Concerns

The resignation gained widespread traction across the tech community after multiple current Anthropic alignment researchers publicly validated Coxon’s concerns on social media, admitting that the industry currently lacks a clear methodology to guarantee control over superintelligent systems:

Evan Hubinger’s Assessment: Evan Hubinger, a lead researcher in Anthropic’s alignment division, endorsed Coxon's core premise, stating he personally estimates a greater than 10% probability that advanced AI could cause human extinction within the next decade. Hubinger added that while Anthropic is making genuine efforts, the industry is "not clearly on track" to solve the alignment problem for superintelligence.
Samuel Marks on Seniority and Fear: Samuel Marks, who leads scalable oversight research at Anthropic, published a personal analysis confirming that catastrophic anxiety is prevalent among frontier AI teams, noting that "in general, the more senior the employee, the more concerned they are" about extinction-level outcomes.

Rejection of "Marketing Stunt" Claims: Coxon explicitly pushed back against critics claiming AI doomsday scenarios are corporate PR tactics, writing that executives frequently soften their messaging publicly while expressing deep concern in private engineering channels.

Political Fallout and Federal Legislative Push

The candid public disclosures immediately reverberated through Capitol Hill, igniting calls from U.S. lawmakers for emergency intervention and legislative restrictions. Senator Bernie Sanders (I-VT) and Representative Greg Casar (D-TX) highlighted the researchers' admissions to promote proposed emergency legislation that would temporarily halt the deployment of unaligned artificial superintelligence and establish a federal oversight agency.

Meanwhile, Senate Commerce Committee Chair Ted Cruz (R-TX) confirmed ongoing bipartisan discussions around regulating catastrophic AI hazards.

President Donald Trump offered a counter-perspective during media remarks, arguing that imposing restrictive domestic mandates could compromise America's strategic position in the global AI race against China.

Anthropic Reaffirms Responsible Scaling Commitments

In response to the public debate, Anthropic defended its risk management framework, pointing to its public Responsible Scaling Policy (RSP) designed to continuously evaluate models for dangerous biological, cyber, and autonomous capabilities. While Anthropic maintained that current operational risk from existing models remains low, the company acknowledged the unprecedented challenges posed by rapid capability scaling. The episode highlights a growing divide within frontier AI labs between commercial scaling incentives and internal alignment warnings as superintelligence timelines contract.