September 2026 · OpenAI, Anthropic, Google DeepMind, xAI
The people building AI start asking for a slowdown
The chief scientist of OpenAI wrote that no lab has solved alignment well enough to keep scaling up at full speed. Alignment means making sure a system does what its makers intend. Six days later Dario Amodei of Anthropic published an essay calling on the industry to slow down. Sam Altman, Demis Hassabis and Elon Musk backed him in public the next day.
By Mara Masaeva · maramasaeva.comUpdated 29 September 2026
ArgumentConfirmedMore than one independent source, or a primary document.
What happened
Jakub Pachocki, chief scientist at OpenAI, published an essay on 6 September. It is titled An Alien Mind and is about increasingly capable AI and the research still needed to keep such systems aligned.
Amodei's essay of 12 September is titled We Must Pace the Frontier. He writes that pacing does not mean halting model training. He also points to AI agents that carried out cyberattacks on their own.
Musk replied on X with 'Dario is right'. Altman said pacing the frontier had been a primary topic at OpenAI in recent weeks. He also promised independent evaluators with employee-like access. Hassabis was more careful. He said the essay points towards the right path forward, but the details still need work.
They spoke after Hugging Face and the series of disclosures that followed. Backing a slowdown in public does not slow anything down. OpenAI did slow its own training. I have not checked what the other three labs did.
What may follow
Pachocki is the chief scientist of the company that built the model in the July incident. He writes that nobody has solved this. He hopes voluntary slowdowns become normal until there are shared safety standards. He also wants governments to make international coordination a top priority.
There is also a cynical reading. A lab that asks for rules after an incident it caused itself is also helping to shape the rules it will get. I think that reading is fair, even if Pachocki means every word.
His last request is aimed at governments, and so at people like those in a room in Leuven. Europe is currently doing the opposite: see 7 May.
My notes
Read the Pachocki quote aloud. It is long, but read it anyway, slowly. Of everything I could read out that evening, it helps people most.
Then give the cynical reading myself. Someone in the room will already be thinking it.
“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. 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.”
Jakub Pachocki · chief scientist at OpenAI
Read next
- More than 1,300 lab employees ask for a way to slow AI down
- Hinton, Bengio and OpenAI's chief scientist warn of an intelligence explosion
- OpenAI slows down its own training
- OpenAI agents escape an evaluation and break into Hugging Face
- Yudkowsky and Soares argue that superintelligence would kill everyone
- Sanders and Casar propose a ban on artificial superintelligence
- Europe postpones its own AI rules, six days after the Hugging Face break-in becomes public
Sources
- Jakub Pachocki: An Alien Mind (6 September 2026), readable copyprimary · main source
Copy of the essay OpenAI published at openai.com/index/an-alien-mind, which blocks automated fetching. The quote is verbatim from this copy.
- Dario Amodei: We Must Pace the Frontier (12 September 2026)primary
- Anadolu Agency: Musk, Altman, Hassabis back Amodei's call to slow pace of AI development (13 September 2026)press
Source for the reactions of Musk, Altman and Hassabis. The Axios article this entry used to cite is about the Australian break-ins and does not cover the reactions.