International

Former Anthropic Researcher Warns AI Could Pose Existential Risk by End of Decade

11 September 2026 21:09 PM

NEWS DESK

AI researcher Jacob Coxon has recently left Anthropic and accused the company and its rival OpenAI of acting irresponsibly in the race to develop increasingly advanced artificial intelligence.

Coxon warned that superhuman AI systems could emerge in the near future and claimed that the technology could potentially pose an existential threat to humanity by the end of the decade.

Coxon, who previously worked at OpenAI, made the remarks in a post on X. His comments have sparked concern over the rapid development of increasingly capable AI systems. By Friday afternoon, the post had received more than 160 million views.

In the post published Wednesday, Coxon said Anthropic and OpenAI were “gambling with human lives” and warned against underestimating the potential risks posed by advanced AI.

He said superhuman AI systems could arrive soon, potentially becoming capable of hacking virtually anything and triggering rapid transformations across different fields. According to Coxon, AI developers genuinely believe that the technology could potentially destroy humanity by the end of the decade.

His comments have revived a long-running debate in Silicon Valley over whether AI can be developed and controlled safely. Anthropic, OpenAI and other companies continue to release increasingly advanced models, while a number of researchers have called for a coordinated slowdown in AI development.

Evan Hubinger, one of Anthropic’s leading researchers in its alignment division, also expressed agreement with Coxon’s assessment in a post on X.

“Jacob is right here,” Hubinger wrote, adding that Anthropic researchers genuinely believe AI could potentially kill everyone. He personally estimated that there is more than a 10% chance of such an outcome occurring within the next decade.

Hubinger also said Anthropic was doing its best but acknowledged that researchers still do not have a solution to the alignment problem associated with superhuman intelligence.

“We don't have a plan for solving the alignment problem for superintelligence,” he said. “It’s not even the case that we’re on a clear path toward a solution.”

In AI research, alignment generally refers to efforts to ensure that AI systems behave in ways consistent with human values, goals and intentions.

Coxon’s comments have also resonated among researchers in the AI industry, particularly those concerned about recursive self-improvement.

The term refers to a hypothetical scenario in which an AI system becomes capable of improving or developing successive versions of itself with little or no direct human intervention.

Companies including Anthropic and OpenAI have warned that systems capable of such rapid self-improvement could increase the risk of humans losing control over advanced AI.

Concerns about AI posing an existential threat to humanity are not new.

In 2023, OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei and other leading AI researchers and executives signed a statement saying that mitigating the risk of human extinction from AI should be a global priority.

In July, around 1,400 AI researchers signed an open letter known as “Pacing the Frontier,” calling on the U.S. government to establish limits and safeguards around the development of increasingly powerful AI systems.

Following the initiative, some members of Congress also took steps to address the rapid advancement of AI.

However, there is still no broad consensus over how advanced AI should be regulated or how quickly the technology should be allowed to progress.

The latest warnings from former and current AI researchers underscore the growing divide within the industry between those seeking to accelerate the development of increasingly capable systems and those calling for stronger safety measures before such technology advances further.