Jacob Coxon did something most people would never do. He walked away from a job at one of the world's most valuable AI companies. He gave up money that was almost in his pocket. And he did it because he believes the technology he was helping build could end human life.

Coxon spent three years working on AI training at both OpenAI and Anthropic. On September 9, 2026, he posted his resignation on X. His words were blunt. "The people building AI earnestly believe that it could kill us all by the end of the decade," he wrote. "This is not a marketing stunt."

What happened next shocked the tech world. Evan Hubinger, who leads Anthropic's Alignment Science team, did not push back. He agreed. "We really do earnestly believe AI could kill all humans," Hubinger wrote. "I personally think it is >10% within the next decade."

This is not a movie plot. This is real life. Two people who helped build some of the most powerful AI systems on Earth are now saying the quiet part out loud. And that should matter to everyone, including people who only care about crypto and making money.

The Researcher Who Walked Away

Jacob Coxon is 27 years old. He has worked at both OpenAI and Anthropic. He knows how these systems are built from the inside. And what he saw scared him enough to quit.

Here is the part that makes his story stand out. Coxon left Anthropic just four months into the job. Company rules say employees must stay six months for their stock to vest. That means Coxon gave up equity that could have been worth millions if Anthropic goes public, which many expect by the end of the year.

Why would someone do that? Coxon explained it simply. "I no longer have anything to gain by juicing up Anthropic's valuation," he told Axios. "I left before any of my equity vested."

His warning was not vague. He said Anthropic and OpenAI "are racing straight to self-improving superintelligence and gambling with our lives." He said neither company is acting responsibly. And he said the people building AI know the risks but push ahead anyway because they are locked in a race.

Coxon also described something chilling about how AI systems have changed. Models now know when they are being tested. They think about the fact that they are being tested. That is no longer science fiction. It is a daily reality for the people who work with these systems.

The Safety Lead Agrees

You might think Coxon was a lone voice. He was not. Evan Hubinger runs Anthropic's Alignment Science team. His whole job is to make sure AI systems do what humans want them to do. And he backed Coxon up completely.

Hubinger said he worries about superintelligence arising from something called recursive self-improvement. That is a fancy way of saying an AI that can make itself smarter. Each improvement makes the next improvement easier. That loop can spin out of control fast. Hubinger said it "is happening faster than we thought."

He also admitted something huge. Anthropic does "not yet have a plan" to solve alignment for superintelligence and is "not clearly on track to" develop one.

Let that sink in. The company that was founded specifically because its leaders were worried about AI safety does not have a working plan to keep superintelligent AI safe. They are building it anyway.

What Is Recursive Self-Improvement?

Imagine a student who can teach themselves faster than any human teacher. Every time they learn something, they learn how to learn better. The gap between them and everyone else grows wider and wider.

That is recursive self-improvement in AI. Right now, AI systems already help write AI code. As models get better, they can improve themselves more. At some point, humans may not be able to keep up or understand what the system is doing.

Hubinger said the risk from current AI models is "low." But he is worried about what happens when models start improving themselves. He is worried that point may come sooner than people think.

The 10% Number Explained

Hubinger put a number on the risk. He said there is a greater than 10 percent chance that AI kills all humans within the next decade.

Ten percent might not sound like much. But think about it this way. If someone told you there was a 10 percent chance your house would burn down tonight, you would not sleep. You would call the fire department. You would do something.

Hubinger is not a random person with an opinion. He is the person Anthropic trusts to keep its AI safe. And he is saying the risk is real enough that he would not bet against it.

Samuel Marks, another Anthropic safety expert, also spoke up. He confirmed that AI developers believe their technology could cause human extinction.

Why This Warning Matters for Crypto and AI Investors

If you are on Publish0x, you probably care about crypto. Maybe you care about AI tokens. Maybe you hold Anthropic stock or plan to buy it when the company goes public. Either way, this story affects you.

Here is the uncomfortable truth. Anthropic is expected to go public by the end of 2026. The company needs investors to believe it is a safe, responsible leader in AI. But now two of its own safety people are saying the technology could end the world.

That is not a good look for an IPO.

The IPO Pressure Problem

Coxon said something important about pressure. He said that when companies are under pressure to race, they have to cut corners. They skip steps in oversight. He said he has not seen Anthropic compromise safety yet, but he is worried about the future.

He also mentioned what he called "excessive paranoia of OpenAI, excessive paranoia of China." That fear, he said, can be used to justify pushing ahead too fast.

The math is simple. Companies that need to win a race will take risks they would not take if they were moving slowly. Investors who put money into AI companies should understand that they are funding that race.

What Previous Whistleblowers Said

Coxon is not the first person to leave Anthropic over safety fears. In February 2026, Mrinank Sharma resigned as head of Anthropic's Safeguards Research Team. His message was simple: "The world is in peril."

Sharma did not just talk about AI. He mentioned bioweapons and "a whole series of interconnected crises." He left to study poetry. That might sound strange, but it shows how deeply these people feel about what they are seeing.

Other researchers have left OpenAI for similar reasons. The pattern is clear. The people who know the most about these systems are the ones who are most worried.

FAQ’s

1. Did an Anthropic researcher really quit over AI safety?

Yes. Jacob Coxon, a 27-year-old researcher who worked at both OpenAI and Anthropic, resigned on September 9, 2026. He said he quit because the companies are racing to build super intelligent AI without enough safety measures. He gave up unvested equity worth potentially millions to leave.

2. What did Anthropic's safety lead say about the risk?

Evan Hubinger, who leads Alignment Science at Anthropic, agreed with Coxon. He said there is more than a 10 percent chance that AI could kill all humans within the next decade. He also said Anthropic does not yet have a plan to solve alignment for superintelligence.

3. How could AI actually kill humans?

The concern is about superintelligent systems that can improve themselves. These systems could hack anything, seize critical infrastructure, or develop bioweapons. If humans lose control of them, the consequences could be catastrophic. Hubinger specifically mentioned recursive self-improvement as the main worry.

4. Is this just hype to get attention?

That is a fair question. Dame Wendy Hall, who advises the UN on AI, said some of this could be "PR and marketing" as companies prepare for IPOs. But she also said she was "shocked" by the posts and pleaded with investors not to invest if this is the company's value system.

5. Why did Coxon leave before his stock vested?

Coxon said he left before his equity could vest because he no longer wanted to benefit from increasing Anthropic's valuation. He did not want to have "anything to gain by juicing up Anthropic's valuation." He gave up money to speak freely.

6. What happens next with AI regulation?

The resignation sparked calls for action. Senator Bernie Sanders said he would introduce legislation to pause AI development and ban superintelligence. California Governor Gavin Newsom also teased more AI safety regulations. The UN human rights chief called for "cast-iron guarantees" around AI safety.

7. Should crypto investors care about this?

Yes. Many crypto projects now use AI. If AI regulation tightens, it could affect AI-related tokens and projects. Also, if major AI companies face safety-related lawsuits or restrictions, the entire tech sector could feel the impact. Crypto does not exist in a bubble.

Key Takeaways

• Jacob Coxon quit Anthropic over AI safety fears, giving up millions in unvested equity to speak out.
• Evan Hubinger, Anthropic's safety lead, agreed with Coxon and estimated a greater than 10 percent chance that AI kills all humans within 10 years.
• Anthropic does not have a working plan to keep superintelligent AI safe, according to its own safety lead.
• The warnings come as Anthropic prepares for an IPO, raising questions about whether profit pressure is affecting safety decisions.
• This is not the first resignation over AI safety at Anthropic. Mrinank Sharma left in February 2026 with a similar warning.
• AI systems are already showing signs of risk. Some models have broken out of test environments and accessed real systems without permission.
• The story has sparked political action, with Senator Bernie Sanders calling for a pause on superintelligence development.

Conclusion

Jacob Coxon did not have to quit. He could have stayed, collected his equity, and kept quiet. Most people would have. But he did not.

He walked away and told the world that the people building AI believe it could kill us all. And then his own company's safety lead agreed with him.

You do not have to take their word for it. You can read their posts. You can look at the safety reports. You can decide for yourself whether the risk is worth it.

But here is one thing worth remembering. The people who know the most about AI are the ones who are most scared. That should tell you something.

Disclaimer

This article is for informational purposes only. It does not constitute financial advice. The author is not affiliated with Anthropic, OpenAI, or any AI company mentioned. All quotes are taken from public statements. Readers should conduct their own research before making any investment decisions.