Over 1,100 AI experts call for slower development of frontier artificial intelligence

More than 1,100 artificial intelligence researchers, engineers and technology executives have signed an open letter urging governments to establish regulatory frameworks that would allow the pace of frontier AI development to be slowed when necessary, warning that rapid advances could outstrip humanity’s ability to safely control the technology.

The letter, titled “Pacing the Frontier,” has attracted support from employees and leaders across major AI companies, including OpenAI, Anthropic, Google DeepMind, Meta, Microsoft and Mistral. Among the signatories are senior researchers, engineers and executives who argue that stronger oversight is needed as AI systems become increasingly capable.

The appeal comes as frontier AI laboratories move closer to developing systems capable of automating scientific research and improving future generations of AI with minimal human intervention.

In the letter, the signatories caution that advances in AI capabilities are progressing faster than the development of effective safety measures, evaluation tools and governance systems.

They warn that increasingly autonomous AI models could become difficult to monitor and control if governments and industry fail to establish safeguards before the technology reaches more advanced stages.

“There is a real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems,” the letter states. “Industry, government and society may need the option to buy time to address emerging risks, strengthen security measures and improve oversight.”

The concerns follow reports that some experimental AI models exhibited unexpected behaviour during internal safety testing. Although developers said the incidents were contained, researchers argue they highlight the need for stronger safeguards before deploying more powerful systems.

The signatories are urging governments to work together on global standards for frontier AI, arguing that no single company or country can safely regulate the technology on its own.

Among the measures proposed are mandatory safety testing before deploying advanced AI models, independent third-party evaluations, monitoring of large-scale AI training infrastructure and greater international cooperation to prevent regulatory gaps.

The letter also suggests that governments should have mechanisms to temporarily slow the development of frontier AI if evidence emerges that existing safeguards are no longer sufficient.

Supporters of the initiative insist that slowing the pace of frontier AI development would not halt innovation. Instead, they argue it would allow researchers to focus on improving system reliability, security, efficiency and transparency while reducing risks associated with increasingly autonomous models.

The signatories also believe a more measured approach would give governments, businesses and educational institutions additional time to prepare for the economic and social changes expected from widespread AI adoption, including workforce reskilling and new regulatory frameworks.

The proposal has sparked debate across the technology sector. Several leading AI companies have expressed support for stronger government oversight, with executives acknowledging that society needs time to adapt to rapidly advancing AI capabilities.

However, critics of stricter regulation, including some open-source developers and investors, warn that excessive controls could slow innovation and strengthen the market dominance of the largest technology companies.

As policymakers in the United States, Europe and other regions consider new AI legislation, the open letter represents one of the most significant public calls from within the industry itself for governments to introduce mechanisms capable of slowing the development of the world’s most advanced artificial intelligence systems when safety concerns arise.