Artificial intelligence researcher Jacob Cookson announced his resignation from Anthropic, directing harsh accusations at it and at OpenAI, and warning that “the frantic race to develop superintelligence capable of improving itself may expose humanity to an unprecedented danger.”

Coxon said through his account on the “X” platform: “I decided to leave Anthropic after I came to the conclusion that the two companies are not behaving responsibly enough, but rather are racing to reach a “superior intelligence” capable of developing itself. This path represents a gamble in our lives.”

He warned against underestimating the power of this technology, anticipating that systems that surpass humans in many fields will soon emerge, capable of penetrating digital systems, and bringing about radical and rapid changes in entire scientific and economic sectors, as well as having actual influence and access to real resources.

Cookson pointed out that “the world has already witnessed a significant acceleration in the capabilities of models within a short period,” and pointed out that “the pace of progress does not show any serious signs of slowing down.”

According to the researcher, “a number of people directly involved in developing these systems seriously believe that they may cause an existential catastrophe for humanity before the end of the current decade.”

On the other hand, Cookson expressed some optimism that efforts could be coordinated between companies and governments. He said, “The recent security incidents related to artificial intelligence systems have made reaching agreements to reduce the speed of the race between American laboratories more realistic.”

But he acknowledged that the world is not currently moving towards preventing a comprehensive international race, especially with the competition between the United States, China and other countries. Containing this race, in Cookson’s opinion, may require politically and economically costly measures, including imposing a temporary ban on developing the capabilities of the most advanced models, until clear regulatory rules and safety guarantees are established.

He made a direct appeal to researchers working inside laboratories, asking them to “think carefully about the coming years and not give in to the idea that the development of superintelligence will happen anyway.”

“Do researchers really want to train a super-intelligent system without having a precise understanding of its mind? Do they continue to work silently under the pressure of competition, or do they use the present moment to demand different, safer conditions?” Cookson asked.