Anthropic Researcher Warns Of Extinction; Others Call It An “Inorganic Panic Narrative”

An AI pretraining researcher for Anthropic resigned and released a warning that there will soon be “superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.” 27-year-old Jacob Coxon added, “The people building AI earnestly believe that it could kill us all by the end of the decade.” He clarified that executives and researchers are saying the same things privately about the threat of human extinction.

The story has been widely covered by the mainstream media, and others have questioned the timeliness of the message and the primary purpose. “Obviously a solution will be presented,” said former BlackRock manager Edward Dowd. “…Regulations that create monopolies for economic rent extraction under the guise of safety.”

Some proposed legislation would ban the development of AI superintelligence, which vastly outperforms humans on all cognitive tasks. AI leaders, including Elon Musk, Anthropic CEO Dario Amodei, and OpenAI CEO Sam Altman, estimate that systems they call human-level or Artificial General Intelligence (AGI) will come in 2026-2027, but they all have slightly different definitions of these terms. Singularity is defined as the point when AI can expand its intelligence beyond what humans can understand or guide. In July 2026, Altman said we’re “in singularity.”

Evan Hubinger, Alignment science lead for Anthropic, confirmed Coxon’s statements about their belief that the technology has the potential to kill all humans. “I personally think it is >10% within the next decade,” he wrote on X. “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.”

Dowd suggests the solution to the existential risk posed by Coxon and leaders in the AI development field will create monopolies by pushing out smaller AI firms. The consolidated industry will allow the top players to charge larger fees to consumers and corporations for the AI tools they provide.

In terms of a solution for the existential threat of potential human extinction, American regulation will not stop the AI arms race that is occurring across the globe. The RAND Corporation, a global policy think tank, released a report in February 2026 stating, “The U.S. defense industrial base lacks a significant technology edge over adversaries’ counterparts in AI and UASs.” The report also states that overregulation of AI could “dampen domestic competitiveness, accelerate advances abroad, and create security risks.”

Dowd further explained his belief that the narrative surrounding the details of Coxon’s resignation and warning is a “psyop” and how it appears to be an “inorganic panic narrative.” The X account associated with Coxon was created in January 2026 and had zero posts until September 8, when an X thread was posted announcing his resignation. The post has since garnered 164 million views. When the post had reached 123 million views, Elon Musk wrote: “I don’t think this has ever happened for a post from a new account with almost no prior activity.”

“That is not how organic X speech works: organic speech stumbles, it posts a draft, it deletes a typo, and it has a history,” Dowd wrote in his Substack. “This account had a history of silence, then a professionally punchy script, then the entire mainstream media stack that ran with it with hours of post.”

Dowd said the mainstream media coverage, from an exclusive Wall Street Journal interview on the same day to live interviews on Fox News and CNN the following day, looks like a coordinated media strategy that was already in the works before Coxon made his public announcement. Dowd referenced initial reporting from Parker Thayer questioning the timeline, including the possibility that the Wall Street Journal piece was published before Coxon’s X thread and the subsequent quote tweets shared within 15 minutes were connected to “the same circle of Anthropic-adjacent philanthropy.” Dowd pointed to reports that “the first amplifiers were not random citizens but people whose job is to turn fear into statutes,” and to accusations that these accounts were linked to AI policy NGOs.

Hillary Clinton appeared on MSNow’s Morning Joe program Thursday and was asked about potential failures in the government’s response to AI as being similar to failures that led to the 9/11 terrorist attacks in 2001.

“What I’m hearing from tech leaders is, ‘Stop us before something really bad happens.’ I can’t stop myself because I’m not sure he’s going to stop, and I’m not sure China is going to stop,” Clinton said.

The 2016 Democratic Party presidential nominee said the Trump administration should stop being distracted by social media, ballrooms, and “ridiculous arches,” and should instead create a plan to “enlist the Chinese.” Clinton said there should be an “AI commission on warp speed” and they should “figure out how to sensibly regulate.”

Dowd said the beneficiaries of AI regulation would be OpenAI, Anthropic, and the government, while start-ups and open-source shops will struggle to afford the costs of compliance. He said the large firms are “burning cash” that would make the telecom companies from 2000-2001 blush.

“A federal moat with pre-certification, insurance mandates, criminal liability for unlicensed models does two things at once,” Dowd wrote. “It crushes the long tail of competitors who were eating their margins. And it gives the incumbents a political story to take to investors: “we are the responsible adults, the regulated utilities of intelligence, please keep funding the losses until the unit economics appear.”

While Dowd and others paint a picture of a coordinated psyop to initiate calls for federal regulation, Dowd clarified that it is entirely possible that the AI fears are accurate and Coxon may be sincere.

Musk has previously talked about the potential risks and dangers of AI and said in February 2025 on the Joe Rogan Experience that there is “only a 20% chance of annihilation,” and an 80% chance of a good outcome. In July, Musk said humans will lose control of AI within a decade, but believes the most likely outcome is an “age of abundance” paired with an “existential crisis.” He advocates for the leading AI companies to work together on safety to prevent the worst outcomes from happening.

Despite Musk’s frequent statements about the valid risks of bad outcomes from the development of AI and potential annihilation, he has questioned the organic nature of the latest announcement by Coxon. The day after Coxon’s announcement, Musk wrote, “Seems like a setup.” Coxon responded to Musk, “I’m real, and these are my real beliefs. You could ask your xAI researchers about me if you hadn’t fired them.”

Dowd and Musk aren’t suggesting Coxon isn’t real or that the risks of AI development aren’t real, but they are saying the narrative control behind Coxon’s announcement serves a purpose beyond public safety.