- Anthropic CEO Dario Amodei outlined a plan to “pace the frontier” of AI development.
- It relies on independent safety evaluators and coordination among labs in democratic countries.
- The proposal follows a former researcher’s high-profile doomsday warning a week earlier.
- Its central weakness is enforceability across competitors and borders.
What Happened
A week after an Anthropic researcher’s doomsday warning rattled the AI world, CEO Dario Amodei outlined his plan to “pace the frontier” of AI development, TechCrunch reported on September 18, 2026. The proposal leans on independent safety evaluators and coordination between AI labs in democratic countries.
Why It Matters
The plan is a direct response to the credibility problem raised when Jacob Coxon, a former pretraining researcher at OpenAI and Anthropic, quit and put the odds of AI catastrophe this decade above ten percent. Amodei is attempting to answer “how do you slow down without unilaterally disarming against rivals” — the question that has stalled every prior pause proposal. Whether it persuades competitors is the test.
Technical Details
The mechanism has two parts: independent evaluators who assess models before deployment, and coordination among labs in democratic countries so that restraint by one does not simply cede ground to another. The hard problems are enforceability and scope — the framework has no obvious hold over labs in non-participating countries, and “independent” evaluation requires access and standards that do not yet exist at scale.
Who’s Affected
Rival labs are being asked to accept external evaluation and coordination they have resisted. Policymakers gain a concrete industry proposal to react to. Safety researchers, including those who have warned from inside the labs, get a test of whether their concerns translate into structural change.
What’s Next
The proposal’s fate depends on whether OpenAI, Google, and others sign on, and whether governments back independent evaluation with authority. Without competitor buy-in and enforcement, “pacing the frontier” risks remaining a principle rather than a practice.
