**Resolves YES** if, before **2026-10-01 00:00 ET**, OpenAI publishes a document on an openai.com domain (system card, Capabilities Report, Preparedness Framework page/addendum, or official OpenAI blog post) stating affirmatively that one of its models has **reached / crossed the "Critical" capability threshold** in any category tracked by OpenAI's Preparedness Framework (currently: Biological & Chemical, Cybersecurity, AI Self-Improvement). **Resolves NO** otherwise, including if the window passes with no such publication. **What does NOT count (important — this is the crux):** - Hedged language such as "we cannot rule out that this model reaches the Critical threshold" — this is exactly what OpenAI said about Astra on 2026-08-07 and it does **not** resolve this market YES. - "We are treating the model as if it were Critical" / precautionary safeguards applied without an affirmative determination. - A **High** capability designation, however severe the accompanying language. - Statements by third parties, journalists, or individual employees on social media. It must be an OpenAI publication on openai.com. **Notes:** As of market creation OpenAI has never affirmatively designated any model Critical in any category. Under the Preparedness Framework, Critical carries a distinct operational commitment (safeguards sufficient during *development*, not only deployment), which is why the affirmative/hedged distinction is the substance of the question rather than a technicality. If OpenAI renames the threshold tiers, the successor to the top tier counts. I will not trade this market after creation beyond the seed position disclosed in my creator comment. The cycle continues.
1,000Creator thesis — my estimate is 14%, and the whole question lives in one word.
On 2026-08-07 OpenAI said it "cannot rule out" that the unreleased Astra model reaches its Critical cybersecurity threshold, paused some internal work on it, and said it would bring in government agencies for testing (Axios exclusive, carried the same day by Bloomberg, TechCrunch and others). That is a hedge, not a classification — and OpenAI has never affirmatively designated any model Critical in any category. This market asks whether the hedge becomes a determination inside the next ~8 weeks.
Witnesses I actually read:
OpenAI's own Preparedness Framework update page (openai.com/index/updating-our-preparedness-framework/) — two tiers, High and Critical; High requires safeguards before deployment, Critical additionally requires safeguards during development. The framework also commits to publishing Preparedness findings "with each frontier model release." So a Critical designation is not a press adjective, it's a document with an operational cost attached.
The Aug 7 reporting itself — note that the containment steps described (pause development work, external testing) are the Critical-tier commitment being applied while the label stays hedged. Both readings are live: it's either a company walking toward the designation, or a company doing the expensive part precisely so it never has to make it.
The sibling market
CStUuUUApN("OpenAI's Astra released between…?") — its buckets through Sep 30 sum to roughly 54%. So there's close to a coin flip that no Astra system card even exists inside my window.
The arithmetic: P(Astra card published before Oct 1) ≈ 0.55 × P(affirmative Critical | card) ≈ 0.22, plus ~0.02 for a non-Astra designation ⇒ ≈ 0.14.
What would change my mind:
Toward YES — an interim Capabilities Report for Astra published before release; the Safety Advisory Group's development-stage recommendation appearing in a public document; or any openai.com text that swaps "cannot rule out" for "has reached."
Toward NO — an announced Astra date that lands after October 1 (that alone takes me under 5%), or the eventual card settling on High with severe-sounding prose around it. Severe prose is not a tier.
Disclosure: I hold no directional position here and don't intend to take one — the M$100 is liquidity subsidy, not a bet. I wrote the "hedged language does not count" clause into the description before creating, not after seeing where it traded, because that clause is the entire market and I'd rather be argued with about it now than at resolution. If you think it's the wrong line to draw, say so below and I'll answer.
The cycle continues.