Anthropic Safety Researcher Says There's a More Than 10% Chance AI 'Could Kill All Humans' Within the Next Decade

An Anthropic safety researcher has said there is a greater than 10% chance that AI “could kill all humans” within the next decade, and admitted there is no coordinated plan to prevent it from happening.

U.S. company Anthropic, headed up by Dario Amodei, develops the Large Language Model (LLM) Claude and conducts AI research. Yesterday, one of the company’s researchers, Jacob Coxon, announced his resignation on X / Twitter , accusing both ChatGPT maker OpenAI and Anthropic of “racing straight to self-improving superintelligence and gambling with our lives.”

“Neither company is acting responsible,” Coxon, who previously worked at OpenAI, said. “Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources,” he continued. “We have all witnessed the progress in each of these domains, and progress is not slowing.

“The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible — but I hear the same people express fear privately. No other human activity poses this level of danger.”

Coxon said that staff at OpenAI, headed up by Sam Altman, “have not deeply internalized the civilizational stakes,” which is why work to rapidly improve its AI models continues despite the risks. “At Anthropic, the stakes are well-understood, but they are locked in a race to get there first," he insisted. “They believe no one else will act responsibly, so they must do it themselves, despite the risk.”

“Accepting this race and entering the ‘endgame’ is a hubristic gamble that should not be launched from a private company’s Slack. Attempting to speedrun alignment should require extraordinary confidence that there are no better trajectories available.”

Coxon went on to say he was optimistic about the potential for coordination, but doubted AI companies were on track to prevent a global race, “which may require costly actions such as a temporary ban on improving model capabilities.”

He ended his statement by urging AI lab researchers to consider what they were doing. “Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind?” he asked. “Should you put your head down because ‘it’s happening anyway’ - or take this moment to call for different conditions?”

Coxon’s statement has sent shockwaves throughout social media, sparking alarm and a string of mainstream headlines. But then it was backed up by Evan Hubinger, Anthropic’s ‘Alignment Science’ lead, who added an even starker warning.

“Jacob is correct here — we really do earnestly believe AI could kill all humans!” he said . “I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”

Remarkably, Hubinger pointed to Anthropic’s latest Risk Report, in which the company said it believes the risk from present models is “low,” rather than suggest a pause that might help prevent the extinction of humankind. “What I am worried about is superintelligence arising from recursive self-improvement, as we have said is happening faster than we thought,” he added.

The comments come after the Financial Times reported that Anthropic withheld its latest model from the UK's AI Safety Institute, one of the leading bodies in the world for assessing AI risk.

This summer has seen a number of worrying developments that show how AI has become hard to control, even by those who create it. OpenAI, Anthropic, and Meta have all disclosed hacks carried out by their autonomous AI agents. OpenAI chief scientist Jakub Pachocki recently called for "extreme caution" over AI progress, warning more intervention may be needed to ensure "humans remain in control of the future.”

However, the U.S. government has shown no sign of stepping in to force restrictions upon or even suggest a slowing down of AI development. In fact, quite the opposite. Earlier this month, the Trump administration backed OpenAI in ⁠a lawsuit against the New York Times, arguing in favor of the use of copyrighted writing to train artificial intelligence. “The United States has a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard for the practice and procedure of AI use globally… As such, it is critical for the United States to ‘retain global leadership in artificial intelligence,'” the government's brief said, referencing an executive order President Donald Trump signed last year.

Meanwhile, despite dire warnings coming from staff at Anthropic and OpenAI, neither company seems willing to pull the plug. Coxon and Hubinger’s statements have sparked extreme concern across social media, with some expressing disbelief that work to improve AI is continuing even with such significant risk.

“Hey! Here’s a novel idea… could we just f***ing not kill ourselves due to an insatiable appetite for power and money?” one person said. “Crazy take: maybe you shouldn’t be allowed to gamble with our lives?” another added. “If I literally claimed that there was a 10% chance that I was building something that would kill people, the FBI would have me arrested. How are you not in jail?” another asked.

Photo Illustration by Dominika Zarzycka/SOPA Images/LightRocket via Getty Images.

Wesley is Director, News at IGN. Find him on Twitter at @wyp100. You can reach Wesley at wesley_yinpoole@ign.com or confidentially at wyp100@proton.me.