A resignation note that traveled fast
Jacob Coxon, an AI researcher who spent three years on pretraining work at OpenAI and Anthropic, announced he was leaving Anthropic. His public note did not stay inside the lab. It moved through developer circles and then into finance chats. Emi Yoshikawa, formerly Ripple’s vice president of global operations and now active in venture capital, noticed the reaction. On X, she pointed to an insider comment saying the post was spreading quickly. The part that got attention: some developers inside Anthropic believe advanced AI could pose a serious threat to humanity in a few years. That is not a small claim. It is also not easy to prove. Still, when people building the models say it, markets listen.
What Coxon claimed
Coxon wrote that neither OpenAI nor Anthropic is acting responsibly. He said both are racing toward self-improving superintelligence and taking risks with human lives. He described a split in how staff at the two companies see the problem. At OpenAI, he suggested, many employees still underestimate the danger. At Anthropic, he said, more people understand the risk but keep moving because they fear losing the race. That fear, according to Coxon, is enough to keep the work going even when the concerns are real. I think that is the uncomfortable part for investors. It turns a technical debate into a management question. If employees doubt the safety story, what does that mean for the company’s next funding round?
Why fintech and VCs are paying attention
Yoshikawa’s post matters because of who she is. She is not a random commentator. She has worked at a major crypto payments company and now sits near investment circles. Her signal suggests that venture funds and fintech firms are starting to treat internal AI lab turmoil as financial risk. A regulatory warning can be modeled. A staff revolt is harder to price. Reputation damage can fade, but it can also affect hiring, partnerships, and valuation talks. For Anthropic and OpenAI, insider criticism before new funding is not just bad press. It can make limited partners ask tougher questions. Some of those questions may not have clean answers.
Call for a pause and international rules
Coxon did not only criticize his former employers. He asked other engineers to think about what they are building. He also called on regulators to create international agreements on the pace of AI development. One idea he raised was a temporary ban on pushing forward the core capabilities of advanced AI. That proposal is unlikely to move quickly. Governments disagree. Companies compete. Researchers change jobs. Even so, the debate has moved from conference panels to resignation letters. That shift could matter more than any single post. The money is watching now. Perhaps the labs should watch back.
