“AI Could Destroy Humanity Within 10 Years,” Former Anthropic Researcher Warns After Resigning

[Financial News] A researcher at U.S. artificial intelligence (AI) company Anthropic has left the company after warning that AI could destroy humanity.

The researcher who issued the warning is 27-year-old Jacob Coxon, who left Anthropic. On the 8th (local time), Coxon posted a message on social media expressing concern about the industry’s race to develop AI systems, drawing widespread attention.
He expressed concern that safety issues could be pushed aside as AI companies compete to develop systems capable of improving their own performance without human intervention. Coxon argued that because AI can improve itself, it could eventually advance beyond human control and reach the point of refusing commands.
Coxon told The Wall Street Journal, “People developing AI genuinely believe that this technology could kill us all within 10 years.” He warned, “Before long, these technologies will become superhuman systems capable of hacking anything, revolutionizing any field overnight, and securing real power and resources.”
He also warned, “Many of the most aggressive scenarios are moving toward becoming reality,” adding, “The situation could already be beyond control by the end of next year.”

Notably, Anthropic executives also agreed with Coxon’s warning. Evan Hubinger, head of Alignment Science at Anthropic, echoed his concerns, saying, “Coxon is right. We genuinely believe that AI could destroy all of humanity.” He added, “Personally, I think the probability will exceed 10% within 10 years.”
Samuel Marks, head of Scalable Oversight, also said, while emphasizing that this was his personal opinion, “AI developers believe their technology could cause human extinction or similarly severe consequences,” adding that this “could happen within the next few years.”
Marks said, “Generally, the higher a person’s position, the greater their concern about this issue.” Nevertheless, he explained that AI development continues because commercial incentives are combined with the belief that companies are competing against developers who could misuse the technology.
Coxon, who is from the United Kingdom, is known to have researched the training of new AI models using massive amounts of data. He left OpenAI and joined Anthropic earlier this year. Although he moved to Anthropic because of its corporate culture, which emphasizes AI safety, Coxon said that the risks of AI ultimately cannot be controlled through the efforts of individual companies alone.
He criticized both OpenAI and Anthropic, saying, “Neither is acting responsibly,” and added, “They are racing toward self-improving superintelligence and gambling with our lives.” He also stressed that government intervention or a coordinated response from the AI industry is necessary to develop artificial general intelligence (AGI) that surpasses human capabilities responsibly.
Earlier, Coxon also joined a statement by AI industry researchers urging governments around the world to establish an international framework capable of regulating the pace of AI development. More than 1,000 researchers signed the statement, including Anthropic CEO Dario Amodei and OpenAI Chief Scientist Jakub Pachocki.
[email protected] Kim Hee-sun Reporter