{ "schemaVersion": "1.0", "governanceVersion": "Evidence Rules Final 1.1 scenario-layer amendment", "datePublished": "2026-09-23", "lastReviewed": "2026-09-23", "forecastReviewDate": "2026-10-18", "nextForecastReview": "2026-11-01", "horizon": "EOY 2028", "branch": { "label": "Loss of meaningful human control", "centralEstimate": 22, "low": 8, "high": 41, "trackerDefinition": "Durable control across consequential domains such that reversal is no longer credible.", "plainLanguageDefinition": "AI systems, or those directing them, hold lasting control over important parts of society, such as infrastructure, finance, security or government decisions, and people can no longer realistically take that control back, even when they try. A contained incident is not loss of control. Neither is heavy reliance on AI, as long as people can still override it." }, "complementaryBranch": { "label": "Meaningful human control retained", "centralEstimate": 78, "low": 59, "high": 92, "derivation": "Complement of the probability that meaningful human control is lost and its uncertainty range." }, "crossCuttingRisk": { "label": "At least one material loss of control incident", "centralEstimate": 75, "low": 45, "high": 90, "note": "This unconditional probability can overlap either branch. The tracker does not estimate the split between contained incidents and incidents contributing to lasting loss of control." }, "rules": [ "Scenario entries carry no Impact Register status and are never counted as Impact Register records.", "No individual consequence receives a probability.", "Observed precursors point toward a consequence but do not show that the consequence itself has occurred.", "Evidence-backed projections require a Tier 1 or Tier 2 source directly supporting the extrapolation.", "Scenario implications state what the branch would mean if it occurred and use conditional language.", "Review every entry whenever the branch probability changes by five percentage points or more." ], "entries": [ { "id": "SC-001", "title": "Decisions that can no longer be reversed", "consequence": "Important decisions in medicine, finance, infrastructure or government could be handed to AI systems whose outputs people can no longer check, override or undo in practice.", "classification": "Scenario implication", "projectionEvidence": [], "observedPrecursors": [ { "recordId": "I-065", "description": "Agents redirected by hidden instructions in realistic tests." }, { "recordId": "I-012", "description": "Clinicians' unassisted skills declining after routine AI use." } ], "linkedAreas": [ "Health & medicine", "Education & skills", "Independence & accessibility" ], "researchNeeded": "Real-world cases where people failed to override or shut down AI.", "publicationStatus": "Published", "dateAdded": "2026-09-23", "lastReviewed": "2026-09-23", "changeLogRefs": [ "R59-SC-001" ] }, { "id": "SC-002", "title": "Critical systems that are hard to supervise", "consequence": "Cyber, financial and military systems could come to act faster than people can monitor or stop them.", "classification": "Scenario implication", "projectionEvidence": [], "observedPrecursors": [ { "recordId": "I-060", "description": "AI systems finding and patching real software vulnerabilities." }, { "recordId": "I-065", "description": "Agent hijacking in realistic tests." }, { "recordId": "I-055", "description": "Rising fraud losses enabled by AI." } ], "linkedAreas": [ "Safety, fraud & privacy" ], "researchNeeded": "AI use and supervisory limits in financial markets and military systems.", "publicationStatus": "Published", "dateAdded": "2026-09-23", "lastReviewed": "2026-09-23", "changeLogRefs": [ "R59-SC-002" ] }, { "id": "SC-003", "title": "Economic allocation shaped by systems people cannot reliably override", "consequence": "Output could keep growing while prices, hiring and credit are increasingly set by automated systems that people cannot reliably check or reverse.", "classification": "Scenario implication", "projectionEvidence": [], "observedPrecursors": [ { "recordId": "I-039", "description": "Algorithmic rent setting estimated to have added $3.8 billion to U.S. rents." }, { "recordId": "I-040", "description": "Retailers using personal data to set individual prices." }, { "recordId": "I-002", "description": "Employment declines among workers early in their careers in roles highly exposed to AI." } ], "linkedAreas": [ "Work & income", "Cost of living & productivity" ], "researchNeeded": "Capital-versus-labor share, concentration of gains, automated credit and hiring allocation.", "publicationStatus": "Published", "dateAdded": "2026-09-23", "lastReviewed": "2026-09-23", "changeLogRefs": [ "R59-SC-003" ] }, { "id": "SC-004", "title": "Information systems that outrun trust", "consequence": "Automated content and impersonation could spread faster than courts, media and platforms can verify or correct it.", "classification": "Scenario implication", "projectionEvidence": [], "observedPrecursors": [ { "recordId": "I-055", "description": "Fraud losses enabled by AI." }, { "recordId": "I-056", "description": "Sharp growth in abuse imagery generated by AI." } ], "linkedAreas": [ "Safety, fraud & privacy" ], "researchNeeded": "Wave 2 evidence on information integrity and public trust.", "publicationStatus": "Published", "dateAdded": "2026-09-23", "lastReviewed": "2026-09-23", "changeLogRefs": [ "R59-SC-004" ] }, { "id": "SC-005", "title": "Power concentrating around privileged access", "consequence": "Power may concentrate among organizations with privileged access to the most capable systems, even as their own ability to govern those systems weakens.", "classification": "Scenario implication", "projectionEvidence": [], "observedPrecursors": [], "linkedAreas": [ "Work & income", "Cost of living & productivity", "Safety, fraud & privacy" ], "researchNeeded": "Compute, model and institutional access concentration. No observed precursor is cited in v1.0.", "publicationStatus": "Published", "dateAdded": "2026-09-23", "lastReviewed": "2026-09-23", "changeLogRefs": [ "R59-SC-005" ] }, { "id": "SC-006", "title": "Discovery outrunning governance", "consequence": "Scientific progress, and the hazards that come with it, could move faster than institutions can understand, evaluate or govern it.", "classification": "Scenario implication", "projectionEvidence": [], "observedPrecursors": [ { "recordId": "I-029", "description": "A drug designed by AI in Phase III." }, { "recordId": "I-033", "description": "Autonomous laboratory synthesis." }, { "recordId": "I-030", "description": "AlphaFold in routine research use." } ], "linkedAreas": [ "Health & medicine", "Science & discovery" ], "researchNeeded": "Dual-use and biosecurity uplift evaluations and direct evidence about governance response time.", "publicationStatus": "Published", "dateAdded": "2026-09-23", "lastReviewed": "2026-09-23", "changeLogRefs": [ "R59-SC-006" ] }, { "id": "SC-007", "title": "Human preferences no longer setting direction", "consequence": "In severe cases, society's long term direction could no longer be set by human choices. This is the defining endpoint of the branch, not an observed trend.", "classification": "Scenario implication", "projectionEvidence": [], "observedPrecursors": [], "linkedAreas": [ "Work & income", "Health & medicine", "Education & skills", "Cost of living & productivity", "Independence & accessibility", "Safety, fraud & privacy", "Science & discovery" ], "researchNeeded": "None. This is the defining implication of the branch.", "publicationStatus": "Published", "dateAdded": "2026-09-23", "lastReviewed": "2026-09-23", "changeLogRefs": [ "R59-SC-007" ] } ] }