A former researcher at Anthropic has raised fresh concerns about the potential dangers of artificial intelligence, warning that increasingly powerful AI systems could eventually pose a serious threat to humanity.
Jacob Coxon, 27, made the comments after leaving the company and sharing a resignation thread on X, where he described the rapid progress being made in AI and the fears held by some people working in the field.
Coxon said future AI systems could become superhuman, with the ability to hack systems, transform entire industries and gain significant access to power and resources.
“These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources,” Coxon wrote.
He added that some people developing the technology genuinely believe it could become dangerous enough to wipe out humanity before the end of the decade.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” he said, stressing that the concern was not simply a publicity strategy.
According to Coxon, some executives and senior researchers may use more cautious language when discussing AI risks publicly, while privately expressing much stronger concerns about where the technology could be heading.
His comments received support from Evan Hubinger, a senior Anthropic employee, who responded that the company’s concerns about AI risks were genuine.
“We really do earnestly believe AI could kill all humans!” Hubinger said.
Hubinger, however, said he personally considers the likelihood of such an outcome to be below 10 per cent over the next decade. He also acknowledged concerns about whether Anthropic currently has a reliable plan to prevent a future AI system from becoming too powerful for humans to control.
He later clarified that Anthropic’s current risk assessment suggests existing AI systems have little chance of acquiring that level of power.
Anthropic CEO Dario Amodei has also repeatedly spoken about the risks associated with the rapid development of artificial intelligence.
Speaking to The New York Times in February, Amodei described AI safety as a complex engineering challenge and warned that something could eventually go wrong with an AI system.
The latest warnings come as pressure grows for governments to introduce stronger safeguards around artificial intelligence. In July, nearly 1,400 employees from AI companies signed an open letter calling on the United States government to regulate the technology and slow its development.
Concerns have also been raised within other leading AI companies. OpenAI chief scientist Jakub Pachocki has warned that AI capabilities are advancing faster than researchers’ ability to reliably monitor and control them.
OpenAI CEO Sam Altman has similarly acknowledged that the pace of AI development may need to be reconsidered to give society enough time to adapt to increasingly powerful systems.
“We may have to pace the rate of AI development to give ourselves enough time for society to harden around these new capability levels,” Altman said.
