Prediction
Answer Options
Resolves YES if at least two of OpenAI, Anthropic, Google, SpaceXAI and Meta announce a pacing deal between them at or after 15 September 2026, 00:04:02 UTC and by 31 December 2026, 11:59 PM ET. A deal first announced before that starting moment does not count, even if it is announced again later. Under that deal the parties must commit to at least one of these undertakings: tying the training, release or deployment of AI models to safety conditions; putting a ceiling on the computing power used to train them; limiting how far or how fast AI systems train or upgrade more capable systems on their own; or otherwise pausing or slowing training or releases. Every lab that is party to the deal must announce it, jointly or separately, through its own official channel or an authorised representative acting in an official capacity. Talk of ongoing negotiations, a deal still to come, or approval of the idea is not enough; acceptance has to be stated outright. A deal covering only a defined class of the parties' models still counts, and so does one brokered by or including a government, provided at least two of those labs are parties to it. It counts even if the deal is not binding, is later altered or withdrawn, or is never put into practice. Official information from the labs themselves is the main basis for the decision; broad agreement among reliable reports may also be used. Otherwise NO.
Each percentage is the market’s current estimate of how likely that answer is. A contract settles at 100p if the answer is right and at 0p if it is wrong.
Reported by versus Markets