Archive · 17 Sep 2026
AI Leaders Split Over Calls for Coordinated Development Slowdown
Some technology leaders worry that very powerful AI systems could become dangerous or difficult for people to control.
They disagree about whether AI companies should slow down together.
Anthropic and OpenAI support more cooperation, testing, monitoring, and safety rules.
Elon Musk and Demis Hassabis also support stronger shared oversight.
Meta’s Mark Zuckerberg says each company should make its systems safe on its own.
NVIDIA’s Jensen Huang says businesses can pursue innovation and safety at the same time without new laws.
Microsoft supports careful development but says many groups, including governments and universities, must be involved.
The debate is about how to keep AI useful and under human control while development continues.
No single approach has won agreement across the industry.
Anthropic CEO Dario Amodei proposed coordinated AI development slowdowns, independent evaluators, and international cooperation.
OpenAI supports pacing development, mandatory safety requirements, independent assessments, and global AI standards.
Meta CEO Mark Zuckerberg and NVIDIA CEO Jensen Huang oppose coordinated slowdowns, favoring companies pacing themselves.
Elon Musk and Google DeepMind co-founder Demis Hassabis backed stronger cooperation, testing, and frontier AI oversight.
Microsoft leaders supported deliberate pacing and embedded evaluators but said AI governance must include governments, academia, and broader society.
- Who
- AI leaders including Dario Amodei, Sam Altman, Mark Zuckerberg, Jensen Huang, Elon Musk, Demis Hassabis, Satya Nadella, and Mustafa Suleyman.
- What
- Technology executives are divided over coordinated slowdowns, regulation, independent evaluations, and other safeguards for advanced AI development.
- When
- The disagreement has intensified recently, with proposals and statements issued this year and in recent months.
- Where
- The debate involves leading AI companies in the United States and calls for cooperation among democratic countries and authoritarian governments, including China.
- Why
- Supporters fear rogue or uncontrolled AI systems could cause serious harm, while opponents say companies can manage safety themselves and that new restrictions could hinder innovation.
Supporters of coordinated safeguards
Supporters of company-led pacing
Whether AI development should slow collectively
Supporters of coordinated safeguards
Anthropic, OpenAI, Elon Musk, and Demis Hassabis support coordinated pacing or stronger cooperation so safety measures can keep up with advanced AI development.
Supporters of company-led pacing
Meta and NVIDIA oppose a coordinated slowdown, arguing that individual companies can determine when their systems are safe enough and that innovation need not be sacrificed.
Role of government and regulation
Supporters of coordinated safeguards
Anthropic and OpenAI support government involvement, national safety requirements, international coordination, and shared standards.
Supporters of company-led pacing
Jensen Huang says existing market incentives are sufficient and that new laws or regulations are unnecessary.
Who should oversee AI safety
Supporters of coordinated safeguards
Supporters favor independent evaluators, competitor testing, and broad cooperation among companies, governments, academia, and countries.
Supporters of company-led pacing
Zuckerberg emphasizes each laboratory’s own responsibility and ability to ensure safety, while some proposals are viewed as placing too much control with a small group of frontier labs.
- Main slowdown proposal
- Anthropic proposed ongoing, employee-like access for independent evaluators to monitor frontier AI companies’ safety practices.
- OpenAI position
- OpenAI supports pacing development, mandatory national safety requirements, independent assessments, incident reporting, and global standards.
- Meta position
- Mark Zuckerberg says each AI laboratory has the responsibility and incentive to pace development safely on its own.
- NVIDIA position
- Jensen Huang argues that market forces are sufficient and that innovation and safety can coexist without new laws or regulations.
- Microsoft position
- Satya Nadella supports deliberate pacing and embedded evaluators but says oversight must include broad representation beyond frontier labs.
- Standards proposal
- Demis Hassabis proposed an industry-funded frontier AI standards body for assessment protocols and national-security testing.
- International challenge
- Amodei called for cooperation among frontier companies and governments, while acknowledging difficulties in coordinating with China.