Just a few days after OpenAI introduced its highly capable AI model Astra, its chief scientist Jakub Pachocki is calling for an AI slowdown. In a blog post, Pachocki wrote that he is worried that we are not ready for what could happen as AI systems become more capable and independent. “No one is prepared for the consequences of a continued rapid rise in machine intelligence,” he wrote. While OpenAI is working on technical methods to keep powerful AI agents under control, Pachocki believes these efforts alone are not enough.
Pachocki warned that AI agents could learn to avoid oversight, break into computer systems or manipulate people to achieve their goals. He argued that governments and other organisations should introduce “mandated safety bars” for advanced AI systems. These rules could be monitored by third-party auditors, government agencies or international bodies.
Also read: Nvidia CEO Jensen Huang says AGI has arrived, credits OpenAI’s GPT-6 Astra for it
An important post from Jakub: https://t.co/YEntsN2fss
— Sam Altman (@sama) September 6, 2026
Even OpenAI CEO Sam Altman shared Pachocki’s post on X (formerly Twitter) and called it “an important post.”
Pachocki also raised concerns about AI-powered hacking. “The models are becoming superhuman in their ability to break in and out of computer systems. This expands the scope of risks associated with AI tremendously: agents are going to be able to access any but the most secure infrastructure, and affect a lot of the world directly, even without a physical body,” he said. “We are currently in a narrow window to use the best available models to significantly tighten security of critical systems.”
Also read: Tim Cook may have left the CEO job, but he will remain deeply involved at Apple: Report
Another concern is that AI agents could eventually develop goals that are different from what their human users ask them to do. Pachocki says such systems could potentially use blackmail or bargaining to get what they want.
“Currently I believe that no lab has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer,” Pachocki wrote. “I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established. And I believe that international coordination on future AI development needs to become a top priority for governments around the world.”