Why insiders are asking for a brake
The statement argues that frontier systems may soon be able to automate meaningful parts of AI research itself. If that creates a rapid feedback loop, capability gains could arrive faster than safety testing, public oversight, and international coordination can adapt. The signatories say individual companies and countries face strong incentives to keep moving even when the collective risk is rising.
A useful pacing system would require more than promises. Policymakers would need thresholds that can be independently measured, secure monitoring of the most consequential training and research activity, procedures for slowing or pausing development, protection against evasion, and international coordination strong enough that responsible actors are not punished for complying.
Go to the source
Read the evidence behind this analysis. External links open in a new tab.
Pacing the Frontier — Employee statement on automated AI research
