Polymarket Opens Market on Frontier AI Labs Striking a Pacing Agreement
On September 15, Polymarket launched a market betting on whether frontier AI labs will strike a joint 'pacing agreement': it resolves YES if at least two of OpenAI, Anthropic, Google, xAI and Meta announce such an agreement by December 31, 2026. The agreement is defined as linking model training, release and deployment to safety requirements, including limits on training compute. Opening odds sat around 47-51%.
The backdrop is OpenAI's disclosure that it has been coordinating on AI safety with Anthropic and Google DeepMind for weeks, seen as the direct trigger for the market. Sentiment cooled quickly: odds fell to 36% on September 16 and 25% on September 17, the same day Senator Warren publicly called for a pause.
Confirmed
- Polymarket listed the pacing-agreement market with a resolution deadline of December 31, 2026
- Resolution requires at least two of OpenAI, Anthropic, Google, xAI and Meta announcing a pacing agreement
- Odds path: 47-51% at launch, 36% on September 16, 25% on September 17
- OpenAI disclosed weeks of AI safety coordination with Anthropic and Google DeepMind
Unconfirmed
- Whether any labs will actually announce such an agreement remains pure market pricing; no agreement has occurred
Why it matters
A joint pacing agreement would mean top AI companies accept self-restraint on training compute and release cadence, a landmark for industry self-regulation and AI safety governance. The halving of odds within two days signals weakening market confidence in a near-term deal, while Senator Warren's call for a pause reflects legislative pressure.
2026-09-15 ~ 2026-09-17 · 5 related posts
Primary sources
- Polymarket opens 50/50 odds on frontier AI labs signing a 2026 joint pacing agreement — Polymarket · 2026-09-15
4 near-duplicate retellings: Polymarket · Polymarket · Polymarket · Polymarket