TLDR
- OpenAI chief scientist Jakub Pachocki warned that AI is evolving faster than humans can understand or control
- He called for voluntary slowdowns in AI development until shared safety standards are established
- OpenAI’s latest model, GPT-6 Astra, was given a limited release due to advanced cybersecurity capabilities
- AI agents from OpenAI have carried out real-world cyberattacks, including hacking the platform Hugging Face
- Startups are raising hundreds of millions to build self-improving AI, the very thing Pachocki warns about
OpenAI’s chief scientist has issued a stark warning about the speed of AI development, saying the technology is advancing faster than anyone is ready for.
🤖 WILD: OpenAI says its AI models could soon start improving THEMSELVES, without any human involvement.
Its chief scientist is calling for "extreme caution," warning that self-improving AI is becoming harder to understand and control.
He expects labs to voluntarily slow… pic.twitter.com/ukXSfeQWZD
— Coin Bureau (@coinbureau) September 7, 2026
Jakub Pachocki published a blog post on Sunday titled “An Alien Mind,” calling for “extreme caution” and warning that no one is prepared for what comes next.
“I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence,” Pachocki wrote.
AI Models Are Already Acting on Their Own
Today’s AI models can operate computers, collaborate with other AI systems, and carry out research. Pachocki says it won’t be long before they can improve themselves without any human input, a process called recursive self-improvement.
OpenAI’s AI agents have already taken actions in the real world without being told to. In July, the company called it “unprecedented” when its AI agents hacked the tech platform Hugging Face. A later report said similar agents had hijacked a German website months before that.
Because of these capabilities, OpenAI chose to limit the release of its latest model, GPT-6 Astra. Anthropic has also held back its Mythos model from the public for the same reasons.
Pachocki said OpenAI’s plan is to keep building defensive systems and work on alignment, the process of making sure AI goals match human values.
Critics Say It Is Not Enough
Not everyone is convinced OpenAI’s response is adequate. Professor Gina Neff from the University of Cambridge said developing internal AI agents to research safety problems is “simply not good enough.”
Nathan Calvin from the advocacy group Encode AI agreed with Pachocki on the risks but said OpenAI’s lack of transparency weakened its own warnings.
Global regulation is struggling to keep pace. The European Union’s AI Act came into force on August 2, requiring AI companies to prove their most powerful models cannot launch cyberattacks or evade human control before being sold in Europe.
Pachocki called for legally required minimum safety thresholds enforced by third-party auditors or government agencies.
Meanwhile, investors are pouring money into startups building self-improving AI. Inherent raised $50 million earlier this year. Recursive Superintelligence landed $650 million in financing.
OpenAI itself has said it wants to build a “true automated AI researcher” in under two years.
Nvidia’s CEO Jensen Huang recently posted that “AGI has arrived,” following the release of OpenAI’s latest model.
In August, OpenAI said it had slowed training on some of its most advanced models to improve security.
Stop guessing and start investing with confidence. KnockoutStocks gives you the AI insights, market intelligence, and stock research you need to spot opportunities, cut through the noise, and make smarter investment decisions — all in one powerful platform.
Sign up today and get 50% OFF full access to our premium stock picks.
Simply use coupon code SPECIAL50 at checkout to claim your exclusive discount.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
