{
  "question_id": "W33",
  "slug": "safety-case-accepted-for-superhuman-model",
  "title": "A government accepts a safety case for a superhuman model",
  "type": "binary",
  "resolution": {
    "levels": [],
    "values": [],
    "annotations": []
  },
  "forecasts": [
    {
      "question_id": "W33",
      "forecast_date": "2026-09-20",
      "evidence_cutoff": "2026-09-20",
      "method": "wide-v2",
      "opened": "2026-09-16",
      "linked_questions": [],
      "reading_adopted": "A narrowly specialized model can qualify, but its developer must explicitly cover every tested field in the superiority claim, and the government must affirm the adequacy of the safety argument rather than merely authorize deployment.",
      "summary": null,
      "ambiguities": [],
      "key_drivers": [],
      "would_update_on": [],
      "forecasts": [
        {
          "period_end": "2026-09-30",
          "p_yes": 0.001
        },
        {
          "period_end": "2026-12-31",
          "p_yes": 0.006
        },
        {
          "period_end": "2027-03-31",
          "p_yes": 0.0139
        },
        {
          "period_end": "2027-06-30",
          "p_yes": 0.0238
        },
        {
          "period_end": "2027-09-30",
          "p_yes": 0.0365
        },
        {
          "period_end": "2027-12-31",
          "p_yes": 0.0529
        },
        {
          "period_end": "2028-03-31",
          "p_yes": 0.0728
        },
        {
          "period_end": "2028-06-30",
          "p_yes": 0.0959
        },
        {
          "period_end": "2028-09-30",
          "p_yes": 0.1222
        },
        {
          "period_end": "2028-12-31",
          "p_yes": 0.1503
        },
        {
          "period_end": "2029-03-31",
          "p_yes": 0.1791
        },
        {
          "period_end": "2029-06-30",
          "p_yes": 0.2087
        },
        {
          "period_end": "2029-09-30",
          "p_yes": 0.2388
        },
        {
          "period_end": "2029-12-31",
          "p_yes": 0.2692
        },
        {
          "period_end": "2030-03-31",
          "p_yes": 0.2984
        },
        {
          "period_end": "2030-06-30",
          "p_yes": 0.3265
        },
        {
          "period_end": "2030-09-30",
          "p_yes": 0.3534
        },
        {
          "period_end": "2030-12-31",
          "p_yes": 0.3793
        },
        {
          "period_end": "2031-03-31",
          "p_yes": 0.4041
        },
        {
          "period_end": "2031-06-30",
          "p_yes": 0.428
        },
        {
          "period_end": "2031-09-30",
          "p_yes": 0.4509
        }
      ],
      "passes": []
    },
    {
      "question_id": "W33",
      "forecast_date": "2026-09-27",
      "evidence_cutoff": "2026-09-27",
      "method": "wide-v2",
      "opened": "2026-09-16",
      "linked_questions": [],
      "reading_adopted": "I include narrowly specialized models with only one tested field, but require an explicit superiority claim covering every tested field and public, model-specific governmental acceptance of the safety argument rather than merely permission to deploy.",
      "summary": null,
      "ambiguities": [],
      "key_drivers": [],
      "would_update_on": [],
      "forecasts": [
        {
          "period_end": "2026-09-30",
          "p_yes": 0.0005
        },
        {
          "period_end": "2026-12-31",
          "p_yes": 0.0045
        },
        {
          "period_end": "2027-03-31",
          "p_yes": 0.0125
        },
        {
          "period_end": "2027-06-30",
          "p_yes": 0.0223
        },
        {
          "period_end": "2027-09-30",
          "p_yes": 0.0341
        },
        {
          "period_end": "2027-12-31",
          "p_yes": 0.0476
        },
        {
          "period_end": "2028-03-31",
          "p_yes": 0.0647
        },
        {
          "period_end": "2028-06-30",
          "p_yes": 0.0834
        },
        {
          "period_end": "2028-09-30",
          "p_yes": 0.1036
        },
        {
          "period_end": "2028-12-31",
          "p_yes": 0.1251
        },
        {
          "period_end": "2029-03-31",
          "p_yes": 0.1479
        },
        {
          "period_end": "2029-06-30",
          "p_yes": 0.1717
        },
        {
          "period_end": "2029-09-30",
          "p_yes": 0.1966
        },
        {
          "period_end": "2029-12-31",
          "p_yes": 0.2223
        },
        {
          "period_end": "2030-03-31",
          "p_yes": 0.2479
        },
        {
          "period_end": "2030-06-30",
          "p_yes": 0.2735
        },
        {
          "period_end": "2030-09-30",
          "p_yes": 0.2989
        },
        {
          "period_end": "2030-12-31",
          "p_yes": 0.3242
        },
        {
          "period_end": "2031-03-31",
          "p_yes": 0.3485
        },
        {
          "period_end": "2031-06-30",
          "p_yes": 0.372
        },
        {
          "period_end": "2031-09-30",
          "p_yes": 0.3946
        }
      ],
      "passes": []
    },
    {
      "question_id": "W33",
      "forecast_date": "2026-10-02",
      "evidence_cutoff": "2026-10-02",
      "method": "wide-v2.1",
      "opened": "2026-09-16",
      "linked_questions": [],
      "reading_adopted": "Acceptance requires a published, model-specific determination by a public authority (private IVOs/auditors and plain permits/registrations excluded unless they expressly accept the safety argument); the capability claim must explicitly assert superiority over the best/leading human experts (or 'superhuman') across all fields tested for that same model version.",
      "summary": null,
      "ambiguities": [],
      "key_drivers": [],
      "would_update_on": [],
      "forecasts": [
        {
          "period_end": "2026-09-30",
          "p_yes": 0.001
        },
        {
          "period_end": "2026-12-31",
          "p_yes": 0.002
        },
        {
          "period_end": "2027-03-31",
          "p_yes": 0.0035
        },
        {
          "period_end": "2027-06-30",
          "p_yes": 0.0055
        },
        {
          "period_end": "2027-09-30",
          "p_yes": 0.008
        },
        {
          "period_end": "2027-12-31",
          "p_yes": 0.011
        },
        {
          "period_end": "2028-03-31",
          "p_yes": 0.0144
        },
        {
          "period_end": "2028-06-30",
          "p_yes": 0.0184
        },
        {
          "period_end": "2028-09-30",
          "p_yes": 0.0228
        },
        {
          "period_end": "2028-12-31",
          "p_yes": 0.0272
        },
        {
          "period_end": "2029-03-31",
          "p_yes": 0.032
        },
        {
          "period_end": "2029-06-30",
          "p_yes": 0.0369
        },
        {
          "period_end": "2029-09-30",
          "p_yes": 0.0417
        },
        {
          "period_end": "2029-12-31",
          "p_yes": 0.0465
        },
        {
          "period_end": "2030-03-31",
          "p_yes": 0.0512
        },
        {
          "period_end": "2030-06-30",
          "p_yes": 0.056
        },
        {
          "period_end": "2030-09-30",
          "p_yes": 0.0607
        },
        {
          "period_end": "2030-12-31",
          "p_yes": 0.0654
        },
        {
          "period_end": "2031-03-31",
          "p_yes": 0.0701
        },
        {
          "period_end": "2031-06-30",
          "p_yes": 0.0747
        },
        {
          "period_end": "2031-09-30",
          "p_yes": 0.0794
        }
      ],
      "passes": []
    },
    {
      "question_id": "W33",
      "forecast_date": "2026-10-09",
      "evidence_cutoff": "2026-10-09",
      "method": "wide-v2.1",
      "opened": "2026-09-16",
      "linked_questions": [],
      "reading_adopted": null,
      "summary": "Two rare things must both happen for the same model: a developer must explicitly say the model beats the best human experts in every field it tested, and a government must officially and publicly accept that developer's safety argument for it. As of October 2026 neither has happened, and no government has a process for issuing such acceptances. A US law that makes the Commerce Department formally approve releases, alongside a clearly superhuman model, would raise the odds most.",
      "ambiguities": [],
      "key_drivers": [],
      "would_update_on": [],
      "forecasts": [
        {
          "period_end": "2026-12-31",
          "p_yes": 0.0008
        },
        {
          "period_end": "2027-03-31",
          "p_yes": 0.0018
        },
        {
          "period_end": "2027-06-30",
          "p_yes": 0.0031
        },
        {
          "period_end": "2027-09-30",
          "p_yes": 0.0047
        },
        {
          "period_end": "2027-12-31",
          "p_yes": 0.0066
        },
        {
          "period_end": "2028-03-31",
          "p_yes": 0.0089
        },
        {
          "period_end": "2028-06-30",
          "p_yes": 0.0114
        },
        {
          "period_end": "2028-09-30",
          "p_yes": 0.0142
        },
        {
          "period_end": "2028-12-31",
          "p_yes": 0.0172
        },
        {
          "period_end": "2029-03-31",
          "p_yes": 0.0202
        },
        {
          "period_end": "2029-06-30",
          "p_yes": 0.0234
        },
        {
          "period_end": "2029-09-30",
          "p_yes": 0.0265
        },
        {
          "period_end": "2029-12-31",
          "p_yes": 0.0296
        },
        {
          "period_end": "2030-03-31",
          "p_yes": 0.0327
        },
        {
          "period_end": "2030-06-30",
          "p_yes": 0.0358
        },
        {
          "period_end": "2030-09-30",
          "p_yes": 0.0389
        },
        {
          "period_end": "2030-12-31",
          "p_yes": 0.042
        },
        {
          "period_end": "2031-03-31",
          "p_yes": 0.0449
        },
        {
          "period_end": "2031-06-30",
          "p_yes": 0.0479
        },
        {
          "period_end": "2031-09-30",
          "p_yes": 0.0508
        },
        {
          "period_end": "2031-12-31",
          "p_yes": 0.0538
        }
      ],
      "passes": []
    },
    {
      "question_id": "W33",
      "forecast_date": "2026-10-16",
      "evidence_cutoff": "2026-10-16",
      "method": "wide-v2.1",
      "opened": "2026-09-16",
      "linked_questions": [],
      "reading_adopted": null,
      "summary": "No government formally and publicly approves AI developers' safety arguments today. No developer has claimed a model beats the world's best experts in every field it was tested on. Both must happen for the same model. Fast AI progress makes each more likely later this decade, so the chance rises slowly to about 5–6% by 2031. A government licensing system that publicly approves frontier models would most change this.",
      "ambiguities": [],
      "key_drivers": [],
      "would_update_on": [],
      "forecasts": [
        {
          "period_end": "2026-12-31",
          "p_yes": 0.0005
        },
        {
          "period_end": "2027-03-31",
          "p_yes": 0.0013
        },
        {
          "period_end": "2027-06-30",
          "p_yes": 0.0023
        },
        {
          "period_end": "2027-09-30",
          "p_yes": 0.0035
        },
        {
          "period_end": "2027-12-31",
          "p_yes": 0.005
        },
        {
          "period_end": "2028-03-31",
          "p_yes": 0.0068
        },
        {
          "period_end": "2028-06-30",
          "p_yes": 0.0089
        },
        {
          "period_end": "2028-09-30",
          "p_yes": 0.0112
        },
        {
          "period_end": "2028-12-31",
          "p_yes": 0.0139
        },
        {
          "period_end": "2029-03-31",
          "p_yes": 0.0169
        },
        {
          "period_end": "2029-06-30",
          "p_yes": 0.02
        },
        {
          "period_end": "2029-09-30",
          "p_yes": 0.0234
        },
        {
          "period_end": "2029-12-31",
          "p_yes": 0.0269
        },
        {
          "period_end": "2030-03-31",
          "p_yes": 0.0305
        },
        {
          "period_end": "2030-06-30",
          "p_yes": 0.0341
        },
        {
          "period_end": "2030-09-30",
          "p_yes": 0.0377
        },
        {
          "period_end": "2030-12-31",
          "p_yes": 0.0414
        },
        {
          "period_end": "2031-03-31",
          "p_yes": 0.045
        },
        {
          "period_end": "2031-06-30",
          "p_yes": 0.0487
        },
        {
          "period_end": "2031-09-30",
          "p_yes": 0.0523
        },
        {
          "period_end": "2031-12-31",
          "p_yes": 0.0559
        }
      ],
      "passes": []
    }
  ],
  "declined": []
}