10th September, 2026

We need to pay attention: AI warning from an Anthropic leader

THIS ARTICLE draws attention to significant warnings issued this week about AI’s power. Evan Hubinger, who is Anthropic’s Alignment Science Lead, wrote, “… we really do earnestly believe AI could kill all humans! 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.” There’s more. Hubinger offered his perspective on the heels of warnings his former colleague Jacob Coxon made on September 8, 2026 in announcing his resignation from Anthropic.

Three weeks ago today, on August 19, 2026, I wrote about AI escaping what’s known as the “sandbox”. In that article, I stated, “We need guardrails in workplaces, and should be advocating for governance/legislation in our respective countries.”

AI is prevalent in all our lives. Like many, I also use GenAI while doing what I can to be aware of AI-related risks. In earlier articles, I’ve written about AI risks that come under headings such as cognitive offloading, copyright issues, data security, legal and reputational risks, “workslop” and so on.

This week, the world has been put on notice about even greater AI risks

– Shelagh Donnelly, September 10, 2026

This week, the world has been put on notice about even greater AI risks. Jacob Coxon, who worked at OpenAI before joining Anthropic, published a series of posts on X/Twitter this Tuesday. He began by writing, “I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.”

Coxon continued, writing, “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. We have all witnessed the progress in each of these domains, and progress is not slowing.” Keep reading for more of what Coxon wrote on September 8, 2026.

• “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.”
• “A common response is ‘if they truly believe this, why are they still building it?’ At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood, but they are locked in a race to get there first – 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.”
• “I am optimistic about the potential for coordination. Warning shots like the Hugging Face attack have made pacing agreements between U.S. labs more viable. I don’t feel like we’re on track to prevent a global race, which may require costly actions such as a temporary ban on improving model capabilities.”
• “If you are a lab researcher, I urge you to consider what the next few years will actually feel like. Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind? Should you put your head down because ‘it’s happening anyway’ – or take this moment to call for different conditions?”

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. We have all witnessed the progress in each of these domains, and progress is not slowing.

– Jacob Coxon, AI researcher who resigned from Anthropic on September 8, 2026

Evan Hubinger responded the same day to Coxon’s posts. Hubinger is Anthropic’s Alignment Science Lead. He wrote, “Jacob is correct here—we really do earnestly believe AI could kill all humans! 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.”

Do I enjoy circulating such dire warnings? I do not. I do encourage you to read more for yourself. You can find Forbes’ piece, Anthropic Alignment Lead Issues Warning About AI Killing Humans As Researcher Resigns, or check Bloomberg News’ article, Some Anthropic Engineers Think AI Might End Us. Why Race Ahead? BBC’s technology reporter Tom Gerken’s piece, Anthropic researcher believes more than 10% chance AI “could kill all humans”, is another example of the headlines circulating in print and digital formats and on television.

Jacob is correct here—we really do earnestly believe AI could kill all humans! 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.

– Evan Hubinger, Anthropic’s Alignment Science Lead, on September 8, 2026

Are we paying sufficient attention to Coxon’s and Hubinger’s concerns? Life is full, even if this didn’t mark the first week of back to school and work routines for many. Others are in the thick of job searches or other significant demands on time and energy. It can be all too easy, when we have a bit of downtime, to turn our minds to less taxing concerns, or to tv, movies or social media scrolling.

That said, we need to pay attention and think for ourselves. Compounding AI-related concerns is the fact that high tech and AI, including the contentious matter of AI data centre construction, represent significant economic factors. I’m not the only person writing about AI this morning; CNBC has published a piece today, Wall Street firm believes the AI stock market boom is “nearing an end.” Here’s why. Even so, we need regulations – and people whose careers are centred around AI are asking for AI governance.

We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.”

– Anthropic leaders Dario Armodei and Jared Kaplan, along with 1,384 employees of “frontier AI companies”; July 2026

Even AI scientists and leaders are asking for governments to establish governance over their products. In Pacing the Frontier, published in July 2026, Anthropic leaders Dario Armodei and Jared Kaplan were among the 1,386 employees of “AI frontier companies” that called for the US government to “support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development”.

If you’d like to add your voice to my annual AI survey, I’m keeping it open a bit longer. This year’s survey includes 12 multiple choice questions, so it should take almost no time for you to add your voice and anonymously document your experience (or lack of) with GenAI.

I’.ve been writing about AI and what I used to refer to digital disruption for years now, including some interviews and articles that now seem rather prescient. I’ve bundled links to some of these pieces on the “AI – Artificial Intelligence” option you’ll see atop your screen. Click here to head to that page, and see how much has changed in the space of less than a decade.

We need guardrails in workplaces, and should be advocating for governance/legislation in our respective countries.

Shelagh Donnelly, August 19, 2026