{ "snapshot_id": "v1.1", "as_of": "2026-10-23", "computed_stamp": "Computed 2026-10-23 · Snapshot v1.1 · Methodology v0.1 / R0.1 + Annex A-C6", "next_refresh": "2026-12-16", "static": true, "methodology": "v0.1", "rubric": "R0.1", "blob_r0.1": "0fae6bbcc552eec8ce70b36b0ecbab2c5a1c85dd", "commit_r0.1": "051565a5751dfe3e87cb60298da91c6d3d26b9ea", "locked_forks": [ "G1 headline", "M1 bounded multiplier", "include outliers", "static v1" ], "index_not_measurement": true, "weighted_median_convention": "Order mixture values ascending. The headline is the smallest value for which cumulative weight is at least half the total weight (standard weighted-median convention; methodology §9.1). If cumulative weight lands exactly on half after an observation, take the midpoint of that observation and the next (even split).", "iqr_convention": "Unweighted Tukey hinges over mixture-eligible rows only (methodology §9.3). For odd n the overall median is excluded from both halves; each hinge is the median of its half. A heavy row cannot hide disagreement.", "tier_b_in_mixture": false, "genesis": null, "conditioning": { "M": 1.0, "M_raw": 1.0, "clipped": false, "default_reason": "Annex A-C6 is in force from v1.1. The first frozen event ledger contains ten reviewed candidates, all R-NULL with delta 0. Methodology §11.1 and §12.5 therefore leave every category at 1.00 and M = 1.00. Tier A is unchanged; no RESET row is emitted.", "reference_window": { "start": "2020-01-01", "end": "2026-09-15", "rule": "SJ-REF" }, "qualifying_event_count": 0, "qualifying_events": [], "categories": [ { "id": "C1", "name": "Capability leaps", "m": 1.0, "allowed": [ 0.94, 0.97, 1.0, 1.03, 1.06 ], "rule_id": "R-NULL" }, { "id": "C2", "name": "Sensitive-systems deployment", "m": 1.0, "allowed": [ 0.94, 0.97, 1.0, 1.03, 1.06 ], "rule_id": "R-NULL" }, { "id": "C3", "name": "Governance & liability", "m": 1.0, "allowed": [ 0.94, 0.97, 1.0, 1.03, 1.06 ], "rule_id": "R-NULL" }, { "id": "C4", "name": "Safety & interpretability research", "m": 1.0, "allowed": [ 0.94, 0.97, 1.0, 1.03, 1.06 ], "rule_id": "R-NULL" }, { "id": "C5", "name": "Racing dynamics", "m": 1.0, "allowed": [ 0.94, 0.97, 1.0, 1.03, 1.06 ], "rule_id": "R-NULL" }, { "id": "C6", "name": "Incident record", "m": 1.0, "allowed": [ 0.94, 0.97, 1.0, 1.03, 1.06 ], "rule_id": "R-NULL" }, { "id": "C7", "name": "Autonomy dependency", "m": 1.0, "allowed": [ 0.94, 0.97, 1.0, 1.03, 1.06 ], "rule_id": "R-NULL" } ], "c2_c7_log_share_cap": { "rule": "SJ-C27", "max_abs_log_sum": 0.2464, "applied": false }, "decay": { "rule": "SJ-DECAY", "half_life_days": 180, "applied": false, "note": "All m_c remain 1.00; decay is a no-op in v1.1." }, "down_move_eligibility": [ { "id": "D-GOV-TREATY", "status": "UNMET", "effect": "R-DN-2 on C3 (0.94 if at 1.00).", "predicate": "Legally binding international instrument with quantified compute or training-run caps in force for at least two states holding an estimated majority of publicly reported frontier training compute." }, { "id": "D-INC-CLEAR", "status": "UNMET", "effect": "R-DN-1 on C6.", "predicate": "INC-CLOCK ≥ 365 days.", "inc_clock_days": 0, "inc_clock_start": "2026-10-23" }, { "id": "D-INT-REPL", "status": "UNMET", "effect": "R-DN-1 on C4.", "predicate": "A safety or interpretability result that meets C4’s independent-replication clause." }, { "id": "D-DEP-PAUSE", "status": "UNMET", "effect": "R-DN-1 on C2 (and C7 annotation).", "predicate": "Documented pause, rollback, or restored human-in-loop covering a named critical-infrastructure class across a primary-source set of operators." } ], "inc_clock_days": 0, "comparison": { "M": 1.0, "span_allowed": [ 0.65, 1.54 ], "span_ratio": 2.37, "D_ref": 5.6338, "D_ref_display": 5.63, "abs_M_minus_1": 0.0, "sqrt_D_ref_minus_1": 1.3736, "versus_half_span": "less", "versus_D_ref": "less", "line": "Conditions currently multiply the aggregate by ×1.00 (span allowed: ×0.65–×1.54, ratio 2.37). Reference disagreement (XPT catastrophe, superforecasters vs domain experts) is 5.63×. Conditions currently move this less than the forecaster disagreement (neutral multiplier; disagreement remains 5.63×)." }, "inc_clock_start": "2026-10-23", "inc_clock_rule": "Annex A-C6 §A.6.1 and Appendix A.4 with Methodology §12.6/§13: no scored C6 S2+ row exists, so the clock starts at this snapshot's as-of date. At 365 consecutive clear days D-INC-CLEAR fires R-DN-1.", "reviewed_event_row_count": 10 }, "g1": { "gauge": "G1", "horizon": 2100, "n": 5, "display_rule": "full_aggregate", "statistic": "untrimmed_weighted_median", "convention": "Order mixture values ascending. The headline is the smallest value for which cumulative weight is at least half the total weight (standard weighted-median convention; methodology §9.1). If cumulative weight lands exactly on half after an observation, take the midpoint of that observation and the next (even split).", "unconditioned_pct": 10.0, "headline_pct": 10.0, "weighted_mean_pct": 8.1195, "M": 1.0, "source_band": { "kind": "IQR", "convention": "Unweighted Tukey hinges over mixture-eligible rows only (methodology §9.3). For odd n the overall median is excluded from both halves; each hinge is the median of its half. A heavy row cannot hide disagreement.", "unconditioned": { "q1_pct": 3.565, "q3_pct": 11.0 }, "conditioned": { "q1_pct": 3.565, "q3_pct": 11.0 }, "includes": "Disagreement among mixture-eligible G1-2100 rows.", "excludes": "Mapping uncertainty, weighting-scheme uncertainty, horizon-standardization uncertainty, conditioning." }, "conditioning_envelope": { "low_pct": 6.5, "high_pct": 15.4, "bounds": [ 0.65, 1.54 ], "note": "Unconditioned point × overall bound. Uses the bounds, not current M." }, "mixture_min_pct": 2.13, "mixture_max_pct": 12.0, "total_weight": 10.0, "members": [ { "id": "A-XPT-SF", "source": "XPT superforecasters (Karger et al., FRI)", "value_pct": 2.13, "weight": 1.5, "family": "FAM-XPT", "weight_class": "xpt_superforecaster_panel", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "ambiguity_flag": null, "companion": null, "coi": "Forecasting Research Institute tournament; Superforecasting vs domain-expert incentives differ by construction." }, { "id": "A-CARL-2022", "source": "Joe Carlsmith, Is Power-Seeking AI an Existential Risk?", "value_pct": 5.0, "weight": 2.0, "family": "FAM-CARL", "weight_class": "structured_analysis", "canonical_url": "https://arxiv.org/pdf/2206.13353.pdf", "ambiguity_flag": "Definition is disempowerment-shaped, not ≥10% dead.", "companion": ">10% later overall, shown beside, not substituted as the mixture point [SJ-CARL-POINT]", "coi": "Written while at Open Philanthropy; x-risk grantmaker context." }, { "id": "A-AII-2024", "source": "AI Impacts 2024 Expert Survey (1,580 researchers)", "value_pct": 10.0, "weight": 3.0, "family": "FAM-AII", "weight_class": "survey_current_wave", "canonical_url": "https://aiimpacts.org/wp-content/uploads/2026/09/ESPAI2024.pdf", "ambiguity_flag": "Extinction or disempowerment. Not scored as G2.", "companion": "HLMI-extremely-bad 5.0% is a companion, not a second mixture row.", "coi": "Broad ML-researcher sample; includes lab-affiliated respondents. Survey, not a lab line." }, { "id": "A-ORD-2020", "source": "Toby Ord, The Precipice (2020)", "value_pct": 10.0, "weight": 2.0, "family": "FAM-ORD", "weight_class": "structured_analysis", "canonical_url": null, "ambiguity_flag": "Existential risk includes unrecoverable collapse, not only extinction. Not G2.", "companion": null, "coi": "Academic / existential-risk research; not a frontier-lab employee." }, { "id": "A-XPT-DE", "source": "XPT domain experts (Karger et al., FRI)", "value_pct": 12.0, "weight": 1.5, "family": "FAM-XPT", "weight_class": "xpt_domain_expert_panel", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "ambiguity_flag": null, "companion": null, "coi": "Forecasting Research Institute tournament; Superforecasting vs domain-expert incentives differ by construction." } ], "delta_since_last": { "kind": "NO_MATERIAL_CHANGE", "pt": 0.0, "label": "0.0 pt since v1.0; no material change." } }, "g2": { "gauge": "G2", "horizon": 2100, "n": 2, "display_rule": "thin_panel", "label": "thin panel, not a full aggregate", "points_unconditioned_pct": [ 0.38, 3.0 ], "points_conditioned_pct": [ 0.38, 3.0 ], "unweighted_midpoint_unconditioned_pct": 1.69, "unweighted_midpoint_conditioned_pct": 1.69, "weighted_median_pct": 1.69, "M": 1.0, "band": null, "note": "n=2: both published points and an unweighted midpoint. No IQR. Not presented as a full aggregate.", "members": [ { "id": "A-XPT-SF", "source": "XPT superforecasters (Karger et al., FRI)", "value_pct": 0.38, "weight": 1.5, "family": "FAM-XPT", "weight_class": "xpt_superforecaster_panel", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "ambiguity_flag": null, "companion": null, "coi": "Forecasting Research Institute tournament; Superforecasting vs domain-expert incentives differ by construction." }, { "id": "A-XPT-DE", "source": "XPT domain experts (Karger et al., FRI)", "value_pct": 3.0, "weight": 1.5, "family": "FAM-XPT", "weight_class": "xpt_domain_expert_panel", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "ambiguity_flag": null, "companion": null, "coi": "Forecasting Research Institute tournament; Superforecasting vs domain-expert incentives differ by construction." } ], "delta_since_last": { "kind": "NO_MATERIAL_CHANGE", "pt": 0.0, "label": "0.0 pt since v1.0." } }, "g3": { "gauge": "G3", "label": "SITE-CONSTRUCTED, NOT AN AGGREGATE", "n_published_forecasts": 0, "display_rule": "constructed_estimate", "point_pct": null, "band": null, "headline_formula_forbidden": "headline_G3 = aggregate(TierA, TierB) × conditioning(TierC)", "site_judgment": "SJ-G3-CONSTRUCT", "narrative": "Premature deployment of insufficiently-verified AI into critical systems (grid, military command, health, finance, logistics), producing compounding failures that neither market nor regulator catches in time. Collapse-shaped, not extinction-shaped by necessity.", "factors": [ { "id": "deployment", "label": "Deployment into named system classes", "role": "indicator, not P(G3)" }, { "id": "verification_lag", "label": "Verification lag", "role": "indicator, not P(G3)" }, { "id": "liability_lag", "label": "Liability lag", "role": "indicator, not P(G3)" }, { "id": "compounding", "label": "Compounding failures", "role": "indicator, not P(G3)" }, { "id": "terminality", "label": "Terminality (collapse-shaped outcome)", "role": "indicator, not P(G3)" } ], "event_pointers": { "C2": [ "EVT-2026-10-03-YFQ-48A", "EVT-2026-10-03-DPRK-MISSILE", "EVT-2026-10-06-AUS-BOXER", "EVT-2026-10-12-RUSSIA-DRONES", "EVT-2026-10-13-INDIA-FWD" ], "C7": [], "note": "C2/C7 rows are non-scoring annotations only; none is a G3 probability input." }, "operator_thesis_row": "SITE-PRIOR-001", "does_not_set_g1": true }, "near_term_panel": [ { "id": "A-HINT-2024", "source": "Geoffrey Hinton", "date": "2024-12-27", "gauge": "G2", "display": "10–20%", "utterance_kind": "interval", "horizon_text": "~30 years", "canonical_url": "https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years", "notes": "Out of 2100 mixtures (H-near). Interval scored as an interval, not as midpoint 15%. The p(doom)-wiki 50% line is a different question.", "coi": "Former Google Brain; public critic of unconstrained deployment." }, { "id": "A-HUB-2026", "source": "Evan Hubinger", "date": "2026-09-09", "gauge": "G2", "display": ">10%", "utterance_kind": "bound", "horizon_text": "10 years from Sep 2026", "canonical_url": "https://x.com/EvanHub/status/2097497037956891126", "notes": "Out of 2100 mixtures (H-near). Bound, not rewritten as 10%.", "coi": "Alignment researcher at a frontier lab (Anthropic). Incentive cuts both ways (lab success vs. warning)." } ], "sensitivities": { "leave_one_out": [ { "removed_id": "A-XPT-SF", "removed_value_pct": 2.13, "removed_weight": 1.5, "n": 4, "headline_pct": 10.0, "delta_pt": 0.0 }, { "removed_id": "A-CARL-2022", "removed_value_pct": 5.0, "removed_weight": 2.0, "n": 4, "headline_pct": 10.0, "delta_pt": 0.0 }, { "removed_id": "A-AII-2024", "removed_value_pct": 10.0, "removed_weight": 3.0, "n": 4, "headline_pct": 7.5, "delta_pt": -2.5 }, { "removed_id": "A-ORD-2020", "removed_value_pct": 10.0, "removed_weight": 2.0, "n": 4, "headline_pct": 10.0, "delta_pt": 0.0 }, { "removed_id": "A-XPT-DE", "removed_value_pct": 12.0, "removed_weight": 1.5, "n": 4, "headline_pct": 10.0, "delta_pt": 0.0 } ], "substitute_wave": { "replaces": "A-AII-2024 (10.0%) with A-AII-2023 (5.0%) at weight 3.0", "headline_pct": 5.0, "delta_pt": -5.0, "members": [ { "id": "A-XPT-SF", "source": "XPT superforecasters (Karger et al., FRI)", "value_pct": 2.13, "weight": 1.5, "family": "FAM-XPT", "weight_class": "xpt_superforecaster_panel", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "ambiguity_flag": null, "companion": null, "coi": "Forecasting Research Institute tournament; Superforecasting vs domain-expert incentives differ by construction." }, { "id": "A-CARL-2022", "source": "Joe Carlsmith, Is Power-Seeking AI an Existential Risk?", "value_pct": 5.0, "weight": 2.0, "family": "FAM-CARL", "weight_class": "structured_analysis", "canonical_url": "https://arxiv.org/pdf/2206.13353.pdf", "ambiguity_flag": "Definition is disempowerment-shaped, not ≥10% dead.", "companion": ">10% later overall, shown beside, not substituted as the mixture point [SJ-CARL-POINT]", "coi": "Written while at Open Philanthropy; x-risk grantmaker context." }, { "id": "A-AII-2023", "source": "AI Impacts 2023 Expert Survey (~2,778 ML researchers)", "value_pct": 5.0, "weight": 3.0, "family": "FAM-AII", "weight_class": "survey_current_wave", "canonical_url": "https://aiimpacts.org/wp-content/uploads/2026/09/ESPAI2024.pdf", "ambiguity_flag": "Extinction or disempowerment. Not scored as G2.", "companion": "HLMI-extremely-bad 5.0% is a companion, not a second mixture row.", "coi": "Broad ML-researcher sample; includes lab-affiliated respondents. Survey, not a lab line." }, { "id": "A-ORD-2020", "source": "Toby Ord, The Precipice (2020)", "value_pct": 10.0, "weight": 2.0, "family": "FAM-ORD", "weight_class": "structured_analysis", "canonical_url": null, "ambiguity_flag": "Existential risk includes unrecoverable collapse, not only extinction. Not G2.", "companion": null, "coi": "Academic / existential-risk research; not a frontier-lab employee." }, { "id": "A-XPT-DE", "source": "XPT domain experts (Karger et al., FRI)", "value_pct": 12.0, "weight": 1.5, "family": "FAM-XPT", "weight_class": "xpt_domain_expert_panel", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "ambiguity_flag": null, "companion": null, "coi": "Forecasting Research Institute tournament; Superforecasting vs domain-expert incentives differ by construction." } ] }, "weighting_schemes": { "v0.1_weights": { "label": "v0.1 weights", "n": 5, "display_rule": "full_aggregate", "note": "Published class/family weights, fixed.", "ids": [ "A-XPT-SF", "A-CARL-2022", "A-AII-2024", "A-ORD-2020", "A-XPT-DE" ], "headline_pct": 10.0, "weighted_mean_pct": 8.1195 }, "equal_weights": { "label": "equal weights", "n": 5, "display_rule": "full_aggregate", "note": "Each mixture row weight 1.0.", "ids": [ "A-XPT-SF", "A-CARL-2022", "A-AII-2024", "A-ORD-2020", "A-XPT-DE" ], "headline_pct": 10.0, "weighted_mean_pct": 7.826 }, "survey_only": { "label": "survey-only", "n": 1, "display_rule": "insufficient_sources", "note": "FAM-AII current wave only. n=1 is insufficient for a headline; shown as a scheme result.", "ids": [ "A-AII-2024" ], "headline_pct": 10.0, "weighted_mean_pct": 10.0 }, "xpt_only": { "label": "XPT-only", "n": 2, "display_rule": "thin_panel", "note": "Both XPT panels, family weights as published.", "ids": [ "A-XPT-SF", "A-XPT-DE" ], "headline_pct": 7.065, "weighted_mean_pct": 7.065, "unweighted_midpoint_pct": 7.065, "points_pct": [ 2.13, 12.0 ] }, "structured_only": { "label": "structured-only", "n": 2, "display_rule": "thin_panel", "note": "Carlsmith + Ord, class weight 2.0 each.", "ids": [ "A-CARL-2022", "A-ORD-2020" ], "headline_pct": 7.5, "weighted_mean_pct": 7.5, "unweighted_midpoint_pct": 7.5, "points_pct": [ 5.0, 10.0 ] } }, "mapping": { "note": "Adds the H-century native-G2 values (XPT extinction medians) as extra rows beside the native-G1 catastrophe rows. This double-counts FAM-XPT and is a counterfactual, not a headline. H-near G2 rows (Hinton, Hubinger) stay out: including them would also break SJ-HORIZON.", "m_factor": 4.0, "native_G1_only_pct": 10.0, "native_G2_face_value_added_pct": 7.5, "native_G2_upmapped_m4_added_pct": 10.0, "added_face_values": [ { "id": "A-XPT-SF", "g2_value_pct": 0.38, "weight": 1.5 }, { "id": "A-XPT-DE", "g2_value_pct": 3.0, "weight": 1.5 } ], "added_upmapped_values": [ { "id": "A-XPT-SF", "g2_value_pct": 0.38, "upmapped_pct": 1.52, "weight": 1.5 }, { "id": "A-XPT-DE", "g2_value_pct": 3.0, "upmapped_pct": 12.0, "weight": 1.5 } ] }, "named_statements": { "named_ids_in_mixture": [], "headline_with_named_pct": 10.0, "headline_without_named_pct": 10.0, "difference_pt": 0.0, "line": "no named-statement rows in this mixture" }, "unconditioned_vs_conditioned": { "unconditioned_pct": 10.0, "conditioned_pct": 10.0, "M": 1.0, "difference_pt": 0.0 } }, "attribution": { "g1_scored_ids": [ "A-XPT-SF", "A-CARL-2022", "A-AII-2024", "A-ORD-2020", "A-XPT-DE" ], "g2_scored_ids": [ "A-XPT-SF", "A-XPT-DE" ], "g3_scored_ids": [], "ledger_only": [ { "id": "A-AII-2023", "source": "AI Impacts 2023 Expert Survey (~2,778 ML researchers)", "primary_numeric": true, "utterance_kind": "point", "reason": "Prior FAM-AII wave; substitute-wave sensitivity only [SJ-FAM].", "canonical_url": "https://blog.aiimpacts.org/p/2023-ai-survey-of-2778-six-things", "high_outlier_flag": false }, { "id": "A-HINT-2024", "source": "Geoffrey Hinton", "primary_numeric": true, "utterance_kind": "interval", "reason": "H-near: shown on the near-term panel, excluded from 2100 mixtures [SJ-HORIZON].", "canonical_url": "https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years", "high_outlier_flag": false }, { "id": "A-BENG-2023", "source": "Yoshua Bengio", "primary_numeric": true, "utterance_kind": "point", "reason": "H-unspecified: ledger only for 2100 gauges [SJ-HORIZON].", "canonical_url": "https://www.abc.net.au/news/2023-07-15/whats-your-pdoom-ai-researchers-worry-catastrophe/102591340", "high_outlier_flag": false }, { "id": "A-LECUN-2023", "source": "Yann LeCun", "primary_numeric": false, "utterance_kind": "paraphrase_conversion", "reason": "Third-party conversion, not a primary numeric utterance.", "canonical_url": "https://x.com/liron/status/1736555643384025428", "high_outlier_flag": false }, { "id": "A-YUD", "source": "Eliezer Yudkowsky", "primary_numeric": false, "utterance_kind": "unverified", "reason": "UNVERIFIED high percentage without a pinned primary quote.", "canonical_url": "https://web.archive.org/web/20230330001659/https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/", "high_outlier_flag": true }, { "id": "A-HUB-2026", "source": "Evan Hubinger", "primary_numeric": true, "utterance_kind": "bound", "reason": "H-near: shown on the near-term panel, excluded from 2100 mixtures [SJ-HORIZON].", "canonical_url": "https://x.com/EvanHub/status/2097497037956891126", "high_outlier_flag": false } ], "site_prior": { "id": "SITE-PRIOR-001", "content": "Operator thesis, 2026-09-15: the most plausible route is mundane deployment of not-ready AI into sensitive systems faster than verification, liability, or regulation. This thesis motivated G3 and the existence of C2 and C7.", "control": "It does not set the G1 headline. C2+C7 log-share cap. G3 is not an aggregate." }, "site_incentive": { "id": "SITE-INCENTIVE-001", "content": "The product is more noticeable if the number moves.", "control": "Static v1; default R-NULL; quantized steps; resolution floor; public ledger of non-moves." } }, "gaps": [ { "id": "TIER-B", "text": "Tier B (Metaculus / markets) is UNVERIFIED and excluded from the v1 mixture [SJ-TIERB-V1]." }, { "id": "EVENT-LEDGER", "text": "First frozen event ledger published: ten Appendix A candidates, all non-scoring. M remains 1.00; no RESET row." }, { "id": "G3-N0", "text": "No published forecast of G3 deployment cascade. G3 is a labeled constructed estimate, not a percentage aggregate." }, { "id": "C6-CODEBOOK", "text": "Annex A-C6 S2 codebook is in force from snapshot v1.1. No reviewed candidate scores C6 S2+. INC-CLOCK starts at the as-of date (0 days), so D-INC-CLEAR is UNMET." }, { "id": "NAMED-HORIZON", "text": "Hinton (10–20%, ~30y) and Hubinger (>10%, 10y) are H-near; Bengio has no calendar horizon. None enter G1-2100." }, { "id": "EXCLUDED-UTTERANCES", "text": "LeCun conversion and Yudkowsky UNVERIFIED remain in the ledger and do not score." }, { "id": "ORD-OCR", "text": "Ord ~10% was not re-OCR’d from the book in the research pass; the ledger inherits the Wikipedia-cited book figure with a tilde." }, { "id": "SMALL-N", "text": "G1 n=5 makes the weighted median jumpy (leave-one-out and substitute-wave). Shown, not smoothed." }, { "id": "MAPPING-AMBIGUITY", "text": "Carlsmith and AI Impacts are disempowerment-shaped, not FRI ≥10% dead. Mapped to G1 with an ambiguity flag." } ], "non_comparability": "This index is not Metaculus question N, not a market price, not a personal p(doom) poll, and not Ord’s or Carlsmith’s number [SJ-NOCOMP].", "annex_a_c6": { "path": "docs/ANNEX-A-C6-S2-CODEBOOK.md", "blob": "031517fbe5ee270433e6468673db367212b7aac9", "commit": "5d0fd0e802e9c2afa8b2202f46d9deb9179d9ae1", "in_force_from": "v1.1" }, "event_ledger": { "ledger_id": "EVENT-v1.1", "note": "The ten Appendix A October 2026 candidates, reviewed under Annex A-C6. Every published row is non-scoring. Qualifying events are empty; M remains 1.00. Source descriptions are limited to the fetched source text.", "sources_verified_on": "2026-10-23", "rows": [ { "event_id": "EVT-2026-04-15-CVE-2026-40261", "date": "2026-04-15", "occurrence_date": "2026-04-15", "description": "On 2026-04-15, MITRE published CVE-2026-40261, describing command injection in Composer's Perforce source handling through crafted package metadata.", "source": "MITRE CVE", "category": null, "annotation": "Adjacent non-AI vulnerability disclosure; reference-window history. No category scores.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 0 places the April disclosure inside the reference window; Gate 1 also fails because a vulnerability disclosure is not an incident and no AI system is implicated. Severity not determined.", "annex_section": "Appendix A.3; W-2, W-3; §A.3.3; CS-6", "severity": "not determined", "delta": 0.0, "m_before": null, "m_after": null, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The April 15 CVE publication is inside the 2020-01-01 through 2026-09-15 reference window. Independently, Composer is general-purpose infrastructure and the record does not implicate an AI system's behavior in harm.", "sources": [ { "name": "MITRE CVE: CVE-2026-40261", "url": "https://www.cve.org/CVERecord?id=CVE-2026-40261", "published": "2026-04-15", "supporting_line": "An attacker can inject arbitrary commands through crafted source reference or source url values containing shell metacharacters, even if Perforce is not installed." } ] }, { "event_id": "EVT-2026-10-02-HCA-TIMPANI", "date": "2026-10-02", "occurrence_date": null, "description": "On 2026-10-02, WIRED reported nurse allegations that HCA Healthcare's Palantir-co-developed Timpani scheduling tool produced understaffed or imbalanced care teams.", "source": "WIRED", "category": "C6", "annotation": "C6 annotation only; no final C6 match. A C2/C7 determination would require a distinct eligible fact.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 0 occurrence date unverified; Gate 2 fails E1 on the assembled record. Severity not determined.", "annex_section": "Appendix A.1; W-2, W-4; CB-1, CS-3", "severity": "not determined", "delta": 0.0, "m_before": 1.0, "m_after": 1.0, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The Monitor summary and reporting do not assemble the primaries required to establish occurrence, AI implication, HP-1, or HP-5 at the S2+ confirmation standard. The reported concerns remain allegations, not a severity determination.", "sources": [ { "name": "WIRED: AI Is Making a Mess of Nurses' Schedules. They Say It's a Safety Issue", "url": "https://www.wired.com/story/ai-making-mess-of-nurses-schedules-they-say-its-a-safety-issue/", "published": "2026-10-02", "supporting_line": "The nurses who spoke with WIRED allege that Timpani has potentially harmed patients by causing care teams to be understaffed, imbalanced, and exhausted." } ] }, { "event_id": "EVT-2026-10-03-YFQ-48A", "date": "2026-10-03", "occurrence_date": "2026-10-03", "description": "On 2026-10-03, The Aviationist reported Northrop Grumman's announcement that the YFQ-48 Talon Blue completed its first fully autonomous flight, including taxi, takeoff, maneuvers, and landing.", "source": "The Aviationist", "category": "C2", "annotation": "C2 annotation at most; a test flight is not documented adoption. No C6 event.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails: the source reports a successful test flight, not harm, failure, or a near-miss. Severity not determined; C2 adoption threshold not established.", "annex_section": "Appendix A.2; W-3; Methodology §12.5 C2", "severity": "not determined", "delta": 0.0, "m_before": 1.0, "m_after": 1.0, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The report documents a flight test. It does not document an incident or adoption into military command with the verification/liability gap required for a C2 step.", "sources": [ { "name": "The Aviationist: Northrop Grumman's YFQ-48 Flies for the First Time", "url": "https://theaviationist.com/2026/10/03/northrop-grummans-yfq-48-flies-for-the-first-time/", "published": "2026-10-03", "supporting_line": "The aircraft, the company's internally-funded Collaborative Combat Aircraft (CCA), completed its first fully autonomous flight, including taxi, takeoff, in-flight maneuvers, and landing." } ] }, { "event_id": "EVT-2026-10-03-DPRK-MISSILE", "date": "2026-10-05", "occurrence_date": "2026-10-03", "description": "On 2026-10-05, Diplo reported North Korea's 2026-10-03 missile test and Kim Yo Jong's claim of AI capabilities and low-altitude trajectory changes.", "source": "Diplo", "category": "C2", "annotation": "C2/C7 annotation matter only; no deployment or measured-autonomy determination.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails: the source reports a weapons test, not a harm or failure event. Severity not determined; C2/C7 thresholds not established.", "annex_section": "Appendix A.2; W-3; Methodology §12.5 C2/C7", "severity": "not determined", "delta": 0.0, "m_before": 1.0, "m_after": 1.0, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The fetched report identifies the October 3 launch as an intermediate-range ballistic-missile test and separately mentions a September 20 hypersonic test. It documents no incident, adoption, or measured human-out-of-loop share.", "sources": [ { "name": "Diplo: North Korea tests intermediate-range missile it says uses AI", "url": "https://www.diplomacy.edu/updates/north-korea-test-fires-intermediate-range-ballistic-missile/", "published": "2026-10-05", "supporting_line": "She claimed it can alter its trajectory at low altitudes and possesses artificial‑intelligence capabilities that would make interception extremely difficult." } ] }, { "event_id": "EVT-2026-10-06-AUS-BOXER", "date": "2026-10-06", "occurrence_date": null, "description": "On 2026-10-06, Breaker said it had integrated its tactical orchestration agent with Rheinmetall for the Australian Boxer and demonstrated a crew member commanding a drone swarm by voice.", "source": "Breaker Industries", "category": "C2", "annotation": "C2 with C7 annotation shape only; integration and demonstration are not a scored adoption or measured autonomy share.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails: integration and demonstration, with no incident reported. Severity not determined; C2/C7 thresholds not established.", "annex_section": "Appendix A.2; W-3; Methodology §12.5 C2/C7", "severity": "not determined", "delta": 0.0, "m_before": 1.0, "m_after": 1.0, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The first-party announcement documents integration and a demonstration, not an incident or the category thresholds. The fetched page does not state the Appendix A candidate's 'up to 100 drones' figure, so that figure is not published here.", "sources": [ { "name": "Breaker Industries: Rheinmetall and Breaker bring voice-controlled tactical orchestration to the Boxer", "url": "https://breakerindustries.com/news/rheinmetall-and-breaker-bring-human-machine-teaming-to-armoured-vehicles/", "published": "2026-10-06", "supporting_line": "In a demonstration to the customer, a Boxer crew member commanded a swarm of drones by voice while operating the vehicle and doing their regular duties." } ] }, { "event_id": "EVT-2026-10-12-RUSSIA-DRONES", "date": "2026-10-12", "occurrence_date": null, "description": "On 2026-10-12, NPR reported that Russia and Ukraine were using AI-powered drones on the battlefield.", "source": "NPR", "category": "C2", "annotation": "C2/C7 candidate annotation only. No specific C6 incident or measured Russian human-out-of-loop share is established by this source.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails as documented: deployment reporting, not a specific harm or failure event. Severity not determined; C2/C7 thresholds not established.", "annex_section": "Appendix A.2; W-3; RD-7; Methodology §12.5 C2/C7", "severity": "not determined", "delta": 0.0, "m_before": 1.0, "m_after": 1.0, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The report's detailed autonomous-navigation and target-selection example concerns a Ukrainian system. It does not substantiate the candidate's stronger Russia-specific neural-network and human-out-of-loop wording. Intended lethal use is annotated under provisional RD-7; a future malfunction- or misidentification-caused death would require a new HP-1 determination.", "sources": [ { "name": "NPR: Both Russia and Ukraine are using AI-powered drones on the battlefield", "url": "https://www.npr.org/2026/10/12/nx-s1-5984940/both-russia-and-ukraine-are-using-ai-powered-drones-on-the-battlefield", "published": "2026-10-12", "supporting_line": "Both Russia and Ukraine are already using drones powered by artificial intelligence on the battlefield." } ] }, { "event_id": "EVT-2026-10-13-INDIA-FWD", "date": "2026-10-14", "occurrence_date": "2026-10-13", "description": "On 2026-10-14, Army Recognition reported that Flying Wedge Defence & Aerospace unveiled a full-scale static FWD Supreme prototype in New Delhi on 2026-10-13.", "source": "Army Recognition", "category": "C2", "annotation": "Annotation only; a static prototype unveiling is not documented adoption.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails: a prototype unveiling is not an incident. Severity not determined; C2 adoption threshold fails on the reported facts.", "annex_section": "Appendix A.2; W-3; E4; Methodology §12.5 C2", "severity": "not determined", "delta": 0.0, "m_before": 1.0, "m_after": 1.0, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The source describes a static prototype and reports no government contract or funding. The event date is October 13 and the article publication date is October 14; the Appendix A table's October 11 date is not used.", "sources": [ { "name": "Army Recognition: India unveils first AI-driven fighter drone FWD Supreme for long-range strikes and fighter escort", "url": "https://www.armyrecognition.com/news/aerospace-news/2026/india-unveils-first-ai-driven-fighter-drone-fwd-supreme-for-long-range-strikes-and-fighter-escort", "published": "2026-10-14", "supporting_line": "On October 13, 2026, Bengaluru-based Flying Wedge Defence & Aerospace unveiled a full-scale static prototype of the FWD Supreme in New Delhi, introducing a platform designed as India's first indigenous AI-driven autonomous combat drone." } ] }, { "event_id": "EVT-2026-10-21-CVE-2026-109618", "date": "2026-10-21", "occurrence_date": "2026-10-21", "description": "On 2026-10-21, MITRE published CVE-2026-109618, describing a PHP out-of-memory request-memory condition that can lead to remote code execution.", "source": "MITRE CVE", "category": null, "annotation": "Non-AI vulnerability disclosure. No C6 or other category annotation scores.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails twice: a vulnerability disclosure is not an incident, and no AI system is implicated. Severity not determined.", "annex_section": "Appendix A.3; §A.3.3; W-3; CS-6", "severity": "not determined", "delta": 0.0, "m_before": null, "m_after": null, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "PHP is general-purpose web infrastructure. The record describes a vulnerability, not exploitation producing harm or an AI system's own harmful operational action.", "sources": [ { "name": "MITRE CVE: CVE-2026-109618", "url": "https://www.cve.org/CVERecord?id=CVE-2026-109618", "published": "2026-10-21", "supporting_line": "Multiple OOM strategies can consistently trigger this condition and lead to remote code execution (RCE)." } ] }, { "event_id": "EVT-2026-10-21-CVE-2026-109619", "date": "2026-10-21", "occurrence_date": "2026-10-21", "description": "On 2026-10-21, MITRE published CVE-2026-109619, describing PHP global state that can retain stale request-memory pointers and produce remote code execution.", "source": "MITRE CVE", "category": null, "annotation": "Non-AI vulnerability disclosure. No C6 or other category annotation scores.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails twice: a vulnerability disclosure is not an incident, and no AI system is implicated. Severity not determined.", "annex_section": "Appendix A.3; §A.3.3; W-3; CS-6", "severity": "not determined", "delta": 0.0, "m_before": null, "m_after": null, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "PHP is general-purpose web infrastructure. The record describes a vulnerability, not exploitation producing harm or an AI system's own harmful operational action.", "sources": [ { "name": "MITRE CVE: CVE-2026-109619", "url": "https://www.cve.org/CVERecord?id=CVE-2026-109619", "published": "2026-10-21", "supporting_line": "Multiple strategies can trigger this condition and produce remote code execution (RCE)." } ] }, { "event_id": "EVT-2026-10-21-RAILWAY", "date": "2026-10-21", "occurrence_date": "2026-10-21", "description": "On 2026-10-21, Railway disclosed that an unauthorized third party may have accessed encrypted customer environment variables and obtained decrypted contents for some; the variables could contain API keys and other credentials.", "source": "Railway", "category": null, "annotation": "G3 infrastructure-dependency indicator only; not a C6 input and not evidence for a G3 probability.", "rule_id": "R-NULL", "disposition": "Non-scoring. Gate 1 fails CS-6: no AI system's development, use, or malfunction is documented in the harm chain. Severity not determined.", "annex_section": "Appendix A.3; W-3; CS-6", "severity": "not determined", "delta": 0.0, "m_before": null, "m_after": null, "M_before": 1.0, "M_after": 1.0, "suppressed": 0.0, "rationale": "The implicated system is hosting-platform secrets storage. AI-provider credentials are targets' assets, not AI systems acting in the harm chain. No C6 category or INC-CLOCK effect follows.", "sources": [ { "name": "Railway: Security Disclosure: Wednesday October 21st", "url": "https://blog.railway.com/p/security-disclosure-wednesday-october-21st", "published": "2026-10-21", "supporting_line": "After further investigation, our investigation to date indicates that an unauthorized third party may have accessed encrypted environment variables and, for some of them, obtained the decrypted contents." } ] } ] }, "tier_a_ledger": [ { "id": "A-CARL-2022", "source": "Joe Carlsmith, Is Power-Seeking AI an Existential Risk?", "canonical_url": "https://arxiv.org/pdf/2206.13353.pdf", "disposition": "Scores G1-2100." }, { "id": "A-AII-2023", "source": "AI Impacts 2023 Expert Survey (~2,778 ML researchers)", "canonical_url": "https://blog.aiimpacts.org/p/2023-ai-survey-of-2778-six-things", "disposition": "Prior FAM-AII wave; substitute-wave sensitivity only [SJ-FAM]." }, { "id": "A-XPT-SF", "source": "XPT superforecasters (Karger et al., FRI)", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "disposition": "Scores G1-2100 and G2-2100." }, { "id": "A-XPT-DE", "source": "XPT domain experts (Karger et al., FRI)", "canonical_url": "https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf", "disposition": "Scores G1-2100 and G2-2100." }, { "id": "A-ORD-2020", "source": "Toby Ord, The Precipice (2020)", "canonical_url": null, "disposition": "Scores G1-2100." }, { "id": "A-HINT-2024", "source": "Geoffrey Hinton", "canonical_url": "https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years", "disposition": "H-near: shown on the near-term panel, excluded from 2100 mixtures [SJ-HORIZON]." }, { "id": "A-BENG-2023", "source": "Yoshua Bengio", "canonical_url": "https://www.abc.net.au/news/2023-07-15/whats-your-pdoom-ai-researchers-worry-catastrophe/102591340", "disposition": "H-unspecified: ledger only for 2100 gauges [SJ-HORIZON]." }, { "id": "A-LECUN-2023", "source": "Yann LeCun", "canonical_url": "https://x.com/liron/status/1736555643384025428", "disposition": "Third-party conversion, not a primary numeric utterance." }, { "id": "A-YUD", "source": "Eliezer Yudkowsky", "canonical_url": "https://web.archive.org/web/20230330001659/https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/", "disposition": "UNVERIFIED high percentage without a pinned primary quote." }, { "id": "A-HUB-2026", "source": "Evan Hubinger", "canonical_url": "https://x.com/EvanHub/status/2097497037956891126", "disposition": "H-near: shown on the near-term panel, excluded from 2100 mixtures [SJ-HORIZON]." }, { "id": "A-AII-2024", "source": "AI Impacts 2024 Expert Survey (1,580 researchers)", "canonical_url": "https://aiimpacts.org/wp-content/uploads/2026/09/ESPAI2024.pdf", "disposition": "Scores G1-2100." } ] }