The chief scientist of OpenAI warned: AI is developing too fast and is difficult to control, calling on the industry to voluntarily slow down if necessary

Recently, several incidents of artificial intelligence (AI) models breaking through or bypassing information security protection mechanisms have been reported, once again raising concerns about the security of AI technology. Jakub Pachocki, chief scientist of OpenAI, which developed the AI ​​model, said that AI is developing rapidly and is becoming increasingly difficult for humans to understand and control. He hopes that AI developers will voluntarily slow down the pace of AI development when necessary for safety reasons.

Bloomberg reported on the 7th that Pahotsky recently wrote an article stating that in a relatively short period of time, AI models have been able to operate computers, collaborate with humans and other AI systems, and execute research projects; and soon, AI may have the ability to improve itself without human intervention, that is, "recursive self-improvement."

He warned that "this is a time for extreme caution" and expressed concern that "no one is truly prepared for the consequences of the continued rapid improvement in machine intelligence capabilities."

Pahotsky believes that in addition to trying to make AI more in line with human interests, developers can also "slow down the pace of future development if necessary." He wrote, "I expect and hope that until a common safety threshold is established, (the industry) voluntarily slowing down will become the norm."

AI models are being introduced at an alarming rate, and breakthroughs have been made in information security and scientific research capabilities. OpenAI stated that when an AI model attacked the research platform Hugging Face a few days ago, it coordinated with each other through message boards without being noticed. OpenAI’s latest model, GPT-6 Astra, has advanced security capabilities, and the company chose to release it on a limited basis; Anthropic has not released its latest Mythos model to the public for similar reasons.

As Pahotsky calls on the industry to slow down the pace of AI development if necessary, more and more startups are investing in the self-improving AI technology he warned about. OpenAI has also stated that it hopes to create "truly automated AI researchers" in less than two years.