A senior researcher at Anthropic has issued a stark public warning about the risks posed by advanced artificial intelligence, estimating a real chance of catastrophic outcomes within the next ten years.
What was said
Evan Hubinger, Anthropic's Alignment Science Lead, posted on X: "We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade." He added that Anthropic is "trying its best" but does not yet have a plan to solve alignment for superintelligence, and is "not clearly on track" to do so. Hubinger later clarified that he believes the risk posed by current AI systems is low, saying his concern centres specifically on the possibility of superintelligence emerging from AI systems that can improve themselves without human involvement, a development he said appears to be progressing faster than expected.
What prompted the comments
Hubinger's remarks followed the resignation of fellow Anthropic researcher Jacob Coxon, who had previously worked on pretraining research at both OpenAI and Anthropic. Coxon accused major AI companies of "racing straight to self-improving superintelligence and gambling with our lives," and argued that industry warnings about existential risk reflect genuine concern rather than a public relations strategy. According to reporting from the Wall Street Journal, Coxon said the most aggressive development timelines could see AI systems become difficult to control as early as the end of 2027.
Coxon also drew a distinction between the two companies' internal cultures, suggesting OpenAI staff had not "deeply internalized the civilizational stakes" involved, while acknowledging that Anthropic employees understand the risks but remain caught in competitive pressure, believing they must move forward because they don't trust other companies to act responsibly instead.
Industry-wide calls for oversight
The comments follow a broader pattern of concern across the AI industry. More than 1,300 employees at AI companies signed an open letter in July called "Pacing the Frontier," urging the US government to support international efforts to develop tools for deliberately pacing the development of advanced AI. Signatories included Anthropic co-founders Dario Amodei and Jared Kaplan, OpenAI Chief Scientist Jakub Pachocki, and Meta AI chief scientist Shengjia Zhao. Pachocki has separately said he doesn't believe any AI lab has yet solved alignment and monitoring sufficiently to continue scaling systems at maximum speed much longer.
In the UK, the AI Security Institute has said it continues working closely with industry partners, including Anthropic, on model safety. In the US, lawmakers have introduced multiple bills addressing AI oversight, including proposals that would allow the government to shut down AI systems considered a threat to the public.
A note on Claude's response
The original article included a quote attributed to Anthropic's Claude chatbot responding to these warnings. I'm not able to verify that this quote is accurate or genuinely reflects how Claude would respond, since I have no way of confirming it came from an actual Claude conversation rather than being written or paraphrased by the article's author. I'd recommend leaving that portion out of your version unless you can verify its source independently.
