Why the People Building AI Want to Slow It Down
Five AI leaders agree that development should slow down before control is lost.
That agreement is the main point of a feature published Sept. 16, 2026, by The Verge titled 'A brief history of AI executives calling for regulation'. It names OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Google DeepMind cofounder Demis Hassabis, Microsoft CEO Satya Nadella and X CEO Elon Musk.
The story starts in January 2015. Musk joined Stephen Hawking in signing an open letter asking for careful AI research. It did not name any specific law, only the request. This came before the advanced systems now called frontier AI were widely sold and used.
The next date is July 2017. Musk told a group of US governors that AI needed rules right away. By that same month, he had put $38 million into OpenAI, according to the same account. The call for limits and the funding existed at the same time.
The broader context here is why this group stands out. These five men do not run the same kind of business. They cover research labs, online platforms and core tech systems. Tech sellers usually say current laws and their own checks are enough. So joint support for slowing down is rare. It fits an old pattern. Builders ask for rules while they are still building. Then regulators, customers and standards groups must decide what proves safety, who checks it, and what happens when tests and safety reports cost time and money.
In my view, the point is not whether these leaders truly mean it. Reasons are often mixed. A lab may want safety rules and also want clear rules that are easier for large, well-funded teams to follow. Both aims can be true at once, along with real worry about losing control. The helpful step is to treat these public calls as one clue and focus on clear, practical detail.
Looking at what this means for working technologists, the real test is not in letters or meetings. It is in written records for each model, repeatable safety tests, slow releases, good logs, clear undo steps, tight control of key model files and training computers, and outside checks. Like car safety checks, these steps show whether talk about slowing down leads to action. If new laws follow this history, those steps become legal requirements, not just good practice.
There is a longer arc worth keeping in mind. I watched my two children grow up as the normal changed, from home computers to phones that are always online to internet services that update on their own. Each change first felt hard to control. Each later got normal rules and tools that made it safer and more routine. AI may take the same path.
For now, the facts are narrow but consistent. The thread runs from a 2015 letter on careful research, through a 2017 call for quick rules, to a 2026 summary where five prominent leaders agree on slowing down before control is lost.
