A growing number of researchers at Anthropic are publicly warning that the race towards increasingly powerful artificial intelligence could create catastrophic risks, including the possibility of human extinction. The extraordinary declarations have intensified the debate over AI safety — while Elon Musk and other critics have questioned whether the warnings are part of a coordinated campaign to encourage tighter regulation.
Researchers break ranks
The controversy accelerated after Anthropic researcher Jacob Coxon resigned, saying that neither Anthropic nor his former employer OpenAI was developing advanced AI responsibly.
Coxon argued that companies were moving towards self-improving superintelligence without having solved the fundamental problem of ensuring that such systems remain under human control.
His warning was subsequently supported publicly by several Anthropic employees. Anna Wang, who works in artificial general intelligence safety and previously worked at Google DeepMind, said there was still no viable scientific solution for managing the risks associated with recursively self-improving AI.
A greater than 10% extinction risk
Perhaps the most striking intervention came from Evan Hubinger, a lead in Anthropic’s alignment research.
Hubinger said he believed there was a greater than 10% probability that AI could cause human extinction within the next decade. Samuel Marks, another Anthropic safety researcher, said concerns about potentially catastrophic outcomes were widespread among AI developers and suggested that more senior employees tended to be more concerned.
These estimates are not scientific forecasts and there is no consensus among AI researchers that artificial intelligence will become capable of causing human extinction. They instead represent individual assessments of uncertain future risks.
Nevertheless, the fact that researchers directly involved in developing frontier AI systems are making such statements has given the debate unusual weight.
Musk calls it a ‘setup’
Elon Musk responded sceptically, describing Coxon’s intervention as appearing to be a “setup” and subsequently suggesting that groundwork for what he called a “psy op” had been established over time.
Other critics promoted theories that the warnings were intended to generate political support for stricter AI regulation.
No substantial evidence has emerged demonstrating that Coxon’s resignation or the supporting statements from Anthropic employees form part of such a coordinated political campaign. Coxon rejected the allegation and insisted that his concerns were genuine.
The dispute is particularly notable because Musk himself has previously warned repeatedly about potential dangers from advanced artificial intelligence, even while his own company develops increasingly capable AI systems.
Anthropic defends its approach
Anthropic says it has consistently acknowledged that advanced AI could create both enormous benefits and unprecedented risks.
The company argues that it is attempting to develop some of the industry’s strongest safeguards while continuing research into how increasingly capable models behave.
The controversy therefore exposes a fundamental tension facing frontier AI companies: the organisations attempting to build the world’s most powerful AI systems are simultaneously warning that those systems could become extremely dangerous.
The threat is not entirely theoretical
Concerns are not limited to hypothetical superintelligence.
Anthropic has also disclosed cases involving attempts to misuse its models for biological research, cyber operations and other potentially dangerous activities. The company said it had identified and disrupted an operation attempting to use its AI technology in work connected with the development of a biological weapon.
AI researcher Gary Marcus has argued that these nearer-term threats deserve at least as much attention as extinction scenarios. His concerns include AI-assisted pathogens, cyberattacks against critical infrastructure and disinformation capable of escalating international conflicts.
A debate moving beyond Silicon Valley
What was once a relatively specialised discussion about AI alignment and “p(doom)” — shorthand for an individual’s estimated probability of catastrophic AI outcomes — is rapidly becoming a political issue.
US lawmakers from different political backgrounds are now discussing stronger oversight, independent safety testing and restrictions on the most advanced systems. The Anthropic researchers’ statements have accelerated those demands.
The central question is becoming increasingly difficult to avoid.
Artificial intelligence promises enormous advances in medicine, science, education, productivity and economic development. Yet some of the people closest to its development are now openly saying that humanity may be advancing faster than its ability to control what it is creating.
Whether their warnings prove prescient or excessively pessimistic remains impossible to know.
But when researchers building frontier AI begin publicly debating the possibility that their own technology could threaten humanity’s survival, the argument over AI safety is no longer confined to science fiction.
It has become part of the global political and economic debate over how far — and how fast — artificial intelligence should be allowed to advance.
Newshub Editorial in North America – 11 September 2026
Ask NF GPT
If you have an account with ChatGPT you get deeper explanations,
background and context related to what you are reading.
Recent Comments