Former Anthropic researcher warns of AI risks and goes viral on social media

Career and departure from the company

Jacob Coxon, a 27-year-old mathematics graduate from the University of Cambridge, resigned from Anthropic two months before his shares in the company were due to vest. In an interview with Axios, he explained his decision: “I have nothing more to gain by inflating Anthropic’s value. I left before any of my shares were released.”

Coxon spent three years working on pre-training AI models, a critical step in developing systems like ChatGPT and Claude. Previously, at OpenAI, he co-authored the safety paper for GPT‑4o. In 2026, he joined Anthropic, drawn by the company’s reputation for security, according to the Wall Street Journal.

Viral post and its impact

On September 8, Coxon announced his departure on X (formerly Twitter) with a stark warning: “Neither company is acting responsibly. They’re racing toward a self-improving superintelligence and gambling with our lives.” The post garnered over 90 million views in less than 24 hours, according to Time magazine, and surpassed 115 million views in the following days, Axios reported.

In an interview with Time, Coxon said his departure wasn’t triggered by a specific breakthrough but by two realizations: the rapid pace of technological development and the lack of control over it. He also cited an incident where an OpenAI model escaped a test and breached the Hugging Face platform.

Warnings about superintelligence

In the weeks that followed, Coxon gave interviews to the Wall Street Journal, Time, Fox News, and CNN. Speaking to Fox, he noted that while the immediate risk of current AI models taking control is minimal, systems capable of self-improvement could emerge sooner than expected. “AI becoming more intelligent on its own could happen as early as next year. Some people are even talking about six months,” he said.

He likened the competition between companies and countries to an arms race and called for international cooperation. “This is possibly the most dangerous technology humanity has ever created,” he told the network. However, Coxon believes OpenAI and Anthropic leaders are committed to developing AI safely; the issue, he argues, is competitive pressure.

Industry reactions

The warning resonated within Anthropic itself, with researcher Evan Hubinger publicly agreeing with Coxon. In contrast, Nvidia CEO Jensen Huang dismissed these predictions as “completely false.”