The First "Covered Frontier Model" Is Publicly Designated Under EO 14409 by the End of 2027

The claim: By December 31, 2027, the US government will have publicly designated at least one AI model a "covered frontier model" under Executive Order 14409 — or a frontier lab will have publicly confirmed that one of its models received the designation. A published assessment methodology alone does not count; the prediction resolves true only when a specific model has actually been run through the process and identified as covered, in a way that is publicly verifiable.

Why I am making it

EO 14409, signed June 2, 2026, creates the covered-frontier-model category but leaves it empty on day one. Before a single designation can exist, the government has to publish assessment criteria, stand up the evaluation capability (reporting points to classified NSA benchmarking for military-relevant models), and have at least one developer engage the voluntary 30-day pre-release access window. The mechanics and the rollout calendar are laid out in the implementation tracker and the full analysis of the order.

I put this at a coin-flip-plus rather than a high-confidence yes for two specific reasons. First, the regime is voluntary, so a designation requires a lab to choose to participate, and the incentive to go first is weak when the category carries security scrutiny and no licensing upside. Second, standing up a defensible, classified-capable assessment methodology inside eighteen months is an aggressive bureaucratic timeline even when the will exists. Against those headwinds, the forces pushing toward a designation are real: a frontier release cadence that all but guarantees qualifying models within the window, clear national-security motivation to demonstrate the regime works, and competitive pressure once any one lab participates publicly. The pull is strong; the voluntary friction is what keeps this at 55 rather than 75.

What would falsify it

If, by December 31, 2027, no model has been publicly designated covered and no lab has confirmed receiving the designation — even if the methodology has been published and the clearinghouse exists — the prediction resolves false. A purely internal or classified designation that never becomes publicly verifiable also counts as a miss for the purposes of this claim, because the resolution condition is public confirmation, not the mere existence of the process.