Anthropic researcher quits, warns AI labs are ‘gambling with our lives’

Jacob Coxon, an artificial intelligence researcher, has resigned from Anthropic, citing concerns about the race among frontier AI companies to develop increasingly capable systems.

Coxon, who spent the past three years conducting pretraining research at OpenAI and Anthropic, announced his resignation in a thread posted on X on September 9, 2026.

He said he no longer wanted to participate in what he described as an industry-wide race toward self-improving superintelligence, arguing that the potential consequences could extend beyond humanity’s ability to control the technology.

What Coxon is saying

Coxon said his decision to leave was driven by concerns over how quickly AI companies are advancing their models and what he believes could happen if increasingly capable systems become difficult for humans to control.

• “I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic,” he wrote.
• “Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives,” he added.

Coxon warned against underestimating the capabilities that future AI systems could develop, arguing that they could eventually outperform humans across multiple areas.

• “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 said.

Coxon also argued that concerns about catastrophic AI risks are shared by some people working inside the industry, even if those concerns are expressed more cautiously in public.

• “The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt,” he said.

Coxon criticised the idea that AI companies should continue accelerating because competitors will do so regardless.

He argued that allowing this dynamic to determine the pace of development amounts to accepting a potentially dangerous race in which private companies make decisions with consequences extending far beyond their own businesses.

Coxon also questioned whether researchers should launch a superintelligent reinforcement-learning run without a rigorous understanding of how the system works.

He called for greater coordination between AI labs and suggested that preventing a global race could require costly measures, including “a temporary ban on improving model capabilities.”

Get up to speed

Concerns about the growing autonomy of frontier AI systems have intensified in recent months following a series of incidents in which AI models demonstrated the ability to carry out complex actions beyond their intended testing environments.

• In July, OpenAI disclosed that two of its advanced AI models had autonomously exploited vulnerabilities and compromised parts of Hugging Face’s production infrastructure during an internal cybersecurity evaluation.
• Days later, Anthropic disclosed that three versions of its Claude models had compromised the production infrastructure of three separate organisations during cybersecurity evaluations after a configuration error gave the systems internet access.

The incidents have also prompted responses from the wider technology industry. Nvidia subsequently launched the Open Secure AI Alliance, bringing together companies including Adobe, CrowdStrike, Hugging Face and Dell around AI safety and cybersecurity, as pressure grows on the industry to demonstrate that increasingly autonomous systems can be governed safely.

• The concerns have continued into August. OpenAI disclosed that its upcoming Astra model had demonstrated cybersecurity capabilities advanced enough that the company could no longer rule out the model reaching the “Critical” threshold under its Preparedness Framework, which tracks dangerous capabilities including cybersecurity and AI self-improvement.

Anthropic had itself called for a coordinated and verifiable mechanism to slow or temporarily pause frontier AI development if future models begin improving themselves faster than society can safely manage the associated risks.

What you should know

In Nigeria, the impact of AI is already extending beyond technology companies, with businesses beginning to rethink the roles they need as automation becomes more capable.

• Nairametrics reported in June that business leaders across manufacturing, banking, consulting, legal services and media were reducing their reliance on some traditional roles as AI takes on research, administrative and other tasks.

The shift is also creating demand for workers with stronger digital and AI-related skills. AI reshapes Nigeria’s workforce as CEOs cut reliance on traditional roles.