1,324 Frontier AI Company Employees Release "Pacing the Frontier" Statement, Urging Government Support for International Coordination

In July 2026, 1,324 employees of frontier AI companies released the "Pacing the Frontier" statement, the core content of which is to request the U.S. government to support international efforts to develop technical and governance tools to deliberately slow the frontier of automated AI development.

The statement notes that leading AI companies believe they may be close to automated AI research, which could push capability development beyond understanding or control. Every company and country faces intense competitive pressure and cannot unilaterally slow down, while the current world lacks tools to deliberately control progress at the frontier. The statement builds on ongoing monitoring efforts of frontier model releases.

Background of the Statement's Signatories and Specific Demands

Signatories include John Schulman, Shengjia Zhao, Ilya Sutskever, and Jasjeet Sekhon. John Schulman commented that the statement helps build common knowledge about the necessity of coordination mechanisms, and he hopes laboratories will voluntarily design these mechanisms before government intervention. Shengjia Zhao noted that the pace of AI progress may outpace society's preparedness, with frontier labs approaching the point of surpassing the best humans on nearly all intellectual metrics, which will bring unprecedented social and safety risks. Ilya Sutskever emphasized that future AI will far exceed anything that exists today, requiring unprecedented measures that must be implemented internationally, otherwise they could be counterproductive. Jasjeet Sekhon's comment mentioned that frontier AI capabilities are advancing rapidly, and evaluation work shows that AI agents are already able to discover and exploit real-world software vulnerabilities.

Competitive Pressure and Mechanism Logic

The statement explains why this is happening: companies and countries face competitive pressure not to unilaterally slow down, while lacking the technical and governance tools to deliberately control frontier progress. Once automated AI research begins, it could rapidly amplify the pace of progress, and society is not yet prepared to handle the resulting systems. The statement does not propose a specific slowdown timeline, but rather focuses on establishing monitoring and coordination mechanisms first.

This logic, combined with confirmed model leak incidents, reveals signs of capability leakage within frontier laboratories, further reinforcing the demand for control tools.

Direct Impact on Various Stakeholders

For developers, the tools required by the statement could increase compliance costs, but may also reduce the risk of project disruptions caused by a loss of capability control. Developers need to adapt to possible future international monitoring requirements based on existing evaluation frameworks such as CyberGym and ExploitGym.

For enterprise users, if the statement translates into policy, it could delay the commercial deployment of the most frontier models, while simultaneously reducing the legal and safety liabilities of deploying insufficiently understood systems. When selecting models, enterprises should prioritize those with publicly disclosed safety evaluation records.

For the upstream and downstream industry chain, hardware and cloud service providers may face slowing demand, while security and governance tool developers gain new market signals. Regulators gain employee-level support, providing bargaining leverage to advance international coordination.

Comparison with Historical Precedents

The "common knowledge building" and "international implementation" requirements mentioned in the statement stand in contrast to past technology governance attempts. Previous nuclear non-proliferation or biosecurity agreements all relied on simultaneous participation by multiple countries, with unilateral action difficult to sustain. The statement explicitly notes that poor implementation could make things worse, consistent with the implementation difficulties brought by hasty agreements in history.

Forward-Looking Assessment

Based on the statement's content, the most likely next step is the U.S. government launching a joint technical working group with the EU, focusing on developing monitoring indicators for automated AI development. Verification signals include whether more laboratory employees join the signing afterward, and whether specific technical monitoring prototypes are made public by the end of 2026.