Resolution criteria

This market resolves to YES if, prior to January 1, 2031, 11:59 PM UTC, a legally binding, indefinite or permanent global ban or moratorium on the development or training of frontier artificial intelligence models comes into force worldwide.

To meet this threshold:

• The restriction must be universal in legal effect (e.g., through an internationally ratified multilateral treaty signed by major AI-developing nations—minimally including the United States, the European Union, and China—or a binding United Nations resolution), prohibiting further training or deployment of frontier AI systems without a set expiration date.

• Voluntary commitments, temporary moratoria with defined sunset dates, localized national bans, or non-binding declarations do not qualify.

• Routine safety evaluations, compliance licensing frameworks, or compute reporting thresholds (such as those under the EU AI Act or executive orders) do not count as a permanent pause.

If no such permanent worldwide ban enters into force before January 1, 2031, this market resolves to NO.

Background

Proposals to halt AI research gained public prominence following the March 2023 open letter from the Future of Life Institute, which called for a temporary six-month pause on training systems more powerful than GPT-4. Since then, prominent AI safety researchers, including Eliezer Yudkowsky, have argued for indefinite, international moratoria to prevent catastrophic existential risks.

In practice, global governance has focused on safety standards, compute thresholds, and alignment benchmarks rather than prohibition. Major international frameworks—such as the Bletchley Declaration, the G7 Hiroshima AI Process, the European Union AI Act, and the United States executive directives—have established oversight mechanisms and voluntary safety pacts, while high-compute frontier model development continues across major jurisdictions.