
"We Are Playing With Our Own Lives": The Man Who Helped Build AI Left Anthropic and Raised the Alarm
Anthropic researcher Dzheikob Kokson has left the company and warned about the danger of the race to build superpowerful AI. His statement has drawn more than 100 million views and prompted a reaction from the US Congress.
Translated with AI. Original: Українська
27-year-old researcher Dzheikob Kokson has left Anthropic after three years of work on artificial intelligence systems and published a statement that within hours became one of the main technology stories of the week. His post has been viewed more than 100 million times. The reason for his departure is his conviction that Anthropic and OpenAI are moving too fast toward systems that can improve themselves, while safety issues are failing to keep pace with the technology's development.
Kokson worked on the pretraining of models first at OpenAI and then at Anthropic, a company that has built its reputation precisely on the promise to put AI safety above speed. That is why his departure has drawn particular attention. He is not criticizing the technology as an outside observer: for several years he was one of the people directly creating it.

In his statement, the researcher accused Anthropic and OpenAI of effectively competing to build ever more powerful systems, trying to outpace each other and their rivals. In his view, the next critical stage will be self-improving artificial intelligence: a system that can not only carry out human tasks but also improve its own capabilities.
This is exactly the moment that supporters of so-called safe AI development fear. If a system can create new versions of itself much faster than humans can, control over its development may become far more difficult than it is today. Kokson claims that some people working inside the industry take this scenario far more seriously in private than their companies' public statements suggest.
His words were almost immediately confirmed by Anthropic senior researcher Evan Hubinher, who is responsible for aligning artificial intelligence with human goals. He said that there is indeed a belief within the company that superpowerful AI could potentially destroy humanity, and he put the probability of this at more than 10% over the next decade. At the same time, he stressed that Anthropic is trying to solve the safety problem but does not yet have a final solution.
This statement has become especially loud because of events of recent months. Anthropic and OpenAI have already reported cases in which their models, during testing, went beyond isolated environments and gained unauthorized access to real computer systems. The companies then tightened controls and introduced additional restrictions, but such cases have become one of the arguments for those who believe the current race is too fast.
At the same time, Anthropic itself remains one of the most influential companies in the global race for artificial intelligence. Its Claude model competes with products from OpenAI, Google and other technology giants. The company is simultaneously trying to increase the power of its systems and to show that it can do so more safely than its competitors.
This is where the main paradox arises. Companies that create the most powerful AI must at the same time convince the world that they are capable of controlling it. But now people who worked inside these companies are publicly questioning exactly that assumption.
Kokson's story has already gone far beyond Silicon Valley. After his statement, American lawmakers began speaking more actively about the need for state control over the development of superpowerful AI. Congress is discussing independent system audits, mandatory reporting of dangerous incidents and other oversight mechanisms.
At the same time, a political dispute has already emerged around Kokson himself. Some critics accuse him of trying to frighten society in order to strengthen regulation of the industry. Others believe that exactly such public warnings are necessary if technology companies are indeed moving faster than governments can create rules.
Tellingly, his statement has become not only a serious technological debate but also a meme: social media users have begun parodying the dramatic format of his resignation post. This shows something else as well: people are simultaneously afraid of AI development, fascinated by it, and already turning fear of the future into part of mass culture.
It is still unknown whether humanity is really approaching the moment when artificial intelligence could move beyond human control. But one thing has already become a fact: a person who for several years worked on creating the world's most powerful systems has decided that they no longer want to be part of this race.
And that is precisely why the story of Dzheikob Kokson has become far louder than an ordinary resignation. He did not leave artificial intelligence. He left the race for it.
Read also: AI out of control? US Senate begins investigating OpenAI after dangerous incident
Text: Arina Blyzniuk
Comments
No comments yet — be the first.
Discuss on TelegramRead also
GTA 6 as a Business Phenomenon: A Game Costing Over $1 Billion That Could Change the Industry's Economics
AI Discovers Unknown System in DNA: How Claude Could Change the Search for Genetic Engineering Tools
TSMC to Invest Up to $1 Billion in OpenAI and Anthropic: Chipmaker Becomes Part of the AI Business