# OpenAI's chief scientist wants the brakes on

> Days after GPT-6 Astra shipped, OpenAI's chief scientist says the whole industry should slow down.
- Title: OpenAI's chief scientist wants the brakes on · TechRadar
- Summary: Days after GPT-6 Astra shipped, OpenAI's chief scientist says the whole industry should slow down. The person building frontier AI is saying nobody…
- Keywords: openai, ai safety, ai policy, regulation, machine intelligence, technology, OpenAI's, chief, scientist, wants, brakes, TechRadar
- Source: TechRadar — https://techradar.com/pro/no-one-is-prepared-for-the-consequences-even-openai-chief-scientist-is-saying-ai-development-needs-to-slow-down
- Published: 2026-09-11T16:05:00+00:00
- Read time: 1 min
- Topics: openai, ai safety, ai policy, regulation, machine intelligence, technology
## The call came from inside
OpenAI Chief Scientist Jakub Pachocki published a post arguing AI development should slow down across the entire industry, not just at his own company.

> no one is prepared for the consequences of a continued rapid rise in machine intelligence
## Timing that raises eyebrows
The post landed just days after OpenAI shipped GPT-6 Astra, its business-focused model, and amid fallout from the Hugging Face incident.
## We are past the line
Pachocki says machine intelligence has started to exceed human capability in ways that reshape things, and the field crossed that line without a manual.

> machine intelligence is starting to exceed that of humans in transformative ways
## Nobody knows how it works
His core worry is not capability but understanding. Deep learning research is still trial and error, so the systems outrun the science explaining them.

> study of deep learning-based AI is largely an experimental science
## AI building the next AI
He points at OpenAI's automated research assistant and recursive self-improvement, where a model helps design its successor, as the reason to choose now.
## Two levers, pull both
Option one: strengthen alignment and monitoring and keep humans in the loop. Option two: coordinate a slowdown. Pachocki says do both, plus government coordination.

> The main levers we have are either steering the process to strengthen alignment and monitoring alongside the AI and find ways to keep people in the loop
## Will anyone actually stop
A slowdown only works if rivals join, and no company wants to be the one that blinked. TechRadar's verdict on the odds is blunt.

> It doesn't seem incredibly likely.
## Key takeaway

The person building frontier AI is saying nobody, including the builders, understands it well enough to keep sprinting.