{
  "id": "S13",
  "slug": "rapid-ai-diffusion-shock",
  "title": "Rapid AI diffusion shock",
  "version": "0.2.0",
  "status": "active",
  "orientation": "downside",
  "family": "D",
  "short_title": "AI diffusion shock",
  "thematics": [
    "ai-cyber-digital",
    "demographics-human-capital",
    "economic-security-supply-chains"
  ],
  "likelihood_type": "systemic",
  "likelihood_note": null,
  "scores": {
    "likelihood_2y": {
      "band": 3,
      "rationale": "Displacement at the 30 per cent exposure level within two years is unlikely; the market-concentration variant is likelier and sits at Extreme."
    },
    "likelihood_10y": {
      "band": 5,
      "rationale": ""
    },
    "impact_systemic": {
      "level": 4,
      "rationale": "On capability grounds rather than GDP loss, which the scenario may raise."
    },
    "impact_national": {
      "level": 4,
      "rationale": ""
    },
    "confidence": {
      "level": "low",
      "reason": "The technology's trajectory is contested."
    },
    "priority": 12,
    "tier": "II",
    "source": "ginc-desk-v0.2"
  },
  "scorecard": {
    "onset": "gradual (years)",
    "duration": "3 years",
    "warning": "months",
    "scope": "global",
    "origin": "accidental (technological)",
    "external_support": "none",
    "policy_response": "current plans",
    "recovery_horizon": "structural"
  },
  "summary": "Frontier AI displaces a large share of knowledge work within three years while models, compute and standards sit with a few foreign providers. Tax bases, graduate labour markets, education systems and the state's own administrative capability are loaded at once. A market-concentration variant covers a sudden repricing of AI-linked assets and the dependency of national digital infrastructure on providers whose decisions are made elsewhere. The Allianz Risk Barometer 2026 ranks AI second (up from tenth); the WEF ranks adverse AI outcomes fifth on the ten-year horizon.",
  "narrative": {
    "dateline": "June 2029",
    "text": "The graduate hiring round that normally takes 40,000 entrants takes 9,000. The accountancy, law and software firms that were the country's largest employers of 22-year-olds now buy the same work from an API billed in a foreign currency. Tax receipts from employment fall faster than the finance ministry's models allowed, because the models were trained on recessions, and this is not one: output is up. The public-service digital team discovers it cannot run its own systems without a provider that changed its terms of service in a quarter. Universities lose a third of applicants to the degrees that paid for the rest. The welfare system is asked to carry a population it was not designed for: qualified, young, unemployed and online. Unions, professional bodies and three political parties converge on a demand for a tax on compute; the compute is abroad. The nations that come through are those with a domestic model capability, a way to tax value where it is consumed, an education system that can retrain at scale and a state that uses the technology better than its citizens' employers do."
  },
  "anchors": [
    {
      "shock_id": "deepseek-r1-release-2025",
      "date": "27 January 2025",
      "event": "DeepSeek R1 release",
      "what_happened": "Nvidia lost about US$590 billion of market value in a day on a cheaper frontier-class model",
      "calibrates": "market-concentration variant"
    },
    {
      "shock_id": "generative-ai-adoption-2023",
      "date": "2023–2026",
      "event": "Generative AI adoption",
      "what_happened": "Fastest consumer technology adoption on record; measurable effects on entry-level hiring in exposed occupations by 2025",
      "calibrates": "diffusion speed"
    },
    {
      "shock_id": "allianz-risk-barometer-2026",
      "date": "January 2026",
      "event": "Allianz Risk Barometer 2026",
      "what_happened": "AI the fastest-rising business risk, second globally",
      "calibrates": "business perception"
    },
    {
      "shock_id": "wef-global-risks-report-2026",
      "date": "January 2026",
      "event": "WEF Global Risks Report 2026",
      "what_happened": "Adverse outcomes of AI fifth on the ten-year horizon, the largest climb between horizons",
      "calibrates": "long-horizon severity"
    },
    {
      "shock_id": "eurasia-group-top-risks-2026",
      "date": "January 2026",
      "event": "Eurasia Group Top Risks 2026",
      "what_happened": "'AI eats its users': extractive business models as a stability risk",
      "calibrates": "political channel"
    },
    {
      "shock_id": "previous-general-purpose-technologies-1990",
      "date": "1990s–2000s",
      "event": "Previous general-purpose technologies",
      "what_happened": "ICT diffusion took two decades to show in productivity; displacement concentrated by region",
      "calibrates": "counter-anchor on speed"
    }
  ],
  "parameters": {
    "common": {
      "severity": {
        "major": "major",
        "severe": "severe",
        "extreme": "extreme"
      },
      "duration": {
        "major": null,
        "severe": "3 years",
        "extreme": null
      },
      "onset": {
        "major": null,
        "severe": "gradual (years)",
        "extreme": null
      },
      "warning": {
        "major": null,
        "severe": "months",
        "extreme": null
      },
      "scope": {
        "major": null,
        "severe": "global",
        "extreme": null
      },
      "origin": {
        "major": null,
        "severe": "accidental (technological)",
        "extreme": null
      },
      "external_support": {
        "major": null,
        "severe": "none",
        "extreme": null
      },
      "concurrency": {
        "major": null,
        "severe": "standalone",
        "extreme": null
      },
      "policy_response": {
        "major": null,
        "severe": "current plans",
        "extreme": null
      },
      "recovery_horizon": {
        "major": null,
        "severe": "structural",
        "extreme": null
      }
    },
    "specific": [
      {
        "name": "Share of occupations with over half of tasks exposed",
        "default": "30 per cent",
        "range": "15–50",
        "unit": "per cent",
        "note": null
      },
      {
        "name": "Diffusion to displacement",
        "default": "3 years",
        "range": "2–7",
        "unit": "years",
        "note": null
      },
      {
        "name": "Domestic frontier-model capability",
        "default": "none",
        "range": "options: some / frontier",
        "unit": null,
        "note": null
      },
      {
        "name": "Compute import dependence",
        "default": "high",
        "range": "low–high",
        "unit": null,
        "note": null
      },
      {
        "name": "Market-concentration shock",
        "default": "off",
        "range": null,
        "unit": null,
        "note": "on at Extreme"
      },
      {
        "name": "Provider terms-of-service shock",
        "default": "on",
        "range": null,
        "unit": null,
        "note": null
      }
    ]
  },
  "transmission": [
    "Task automation in exposed occupations.",
    "Entry-level hiring collapse before mid-career displacement.",
    "Tax-base erosion and welfare demand.",
    "Education demand shifts; institutions destabilised.",
    "Dependency on foreign providers for public and private systems.",
    "Political demands for taxation and restriction; capital and talent respond.",
    "Asset repricing if the economics of the providers change."
  ],
  "loading": [
    {
      "dimension": "Hard",
      "domain": "defence-security",
      "load": "Medium",
      "channel": "dependence on foreign models for defence and intelligence functions"
    },
    {
      "dimension": "Hard",
      "domain": "strategic-infrastructure",
      "load": "Medium",
      "channel": "data centres, power demand, connectivity"
    },
    {
      "dimension": "Hard",
      "domain": "critical-technology",
      "load": "High",
      "channel": "domestic model, compute and data capability"
    },
    {
      "dimension": "Soft",
      "domain": "government-effectiveness",
      "load": "High",
      "channel": "tax design, regulation, state use of AI, procurement dependence"
    },
    {
      "dimension": "Soft",
      "domain": "human-capital",
      "load": "High",
      "channel": "graduate labour market, retraining, education finance"
    },
    {
      "dimension": "Soft",
      "domain": "influence-cohesion",
      "load": "High",
      "channel": "distributional conflict, information environment, standards influence"
    },
    {
      "dimension": "Economic",
      "domain": "macro-financial",
      "load": "High",
      "channel": "tax base, asset concentration, productivity"
    },
    {
      "dimension": "Economic",
      "domain": "industry-trade-supply",
      "load": "Medium",
      "channel": "services trade, firm concentration"
    },
    {
      "dimension": "Economic",
      "domain": "energy-resources",
      "load": "Medium",
      "channel": "compute power demand"
    }
  ],
  "stakeholders": {
    "government": {
      "exposure": "Tax base, education and welfare design, procurement dependence",
      "actions": [
        "Model exposure by occupation and tax line",
        "Build sovereign fallbacks for critical public systems",
        "Fund retraining at scale",
        "Negotiate standards"
      ],
      "watch": [
        "Entry-level vacancy data",
        "AI share of public procurement"
      ],
      "relevance": 4
    },
    "technology": {
      "exposure": "The shock is also the opportunity",
      "actions": [
        "Diversify model providers",
        "Invest in domestic capability where it is strategic"
      ],
      "watch": [
        "Provider pricing and terms changes"
      ],
      "relevance": 4
    },
    "investors": {
      "exposure": "Equity concentration, labour-intensive services, education",
      "actions": [
        "Scenario on a 30 per cent occupational exposure and a DeepSeek-style repricing"
      ],
      "watch": [
        "Hyperscaler capex",
        "Model price curves"
      ],
      "relevance": 4
    },
    "public": {
      "exposure": "Careers and education choices",
      "actions": [],
      "watch": [
        "Occupational exposure data"
      ],
      "relevance": 5
    }
  },
  "regions": [
    {
      "region": "north-america",
      "exposure": "High",
      "rationale": "Origin; concentration and displacement"
    },
    {
      "region": "europe",
      "exposure": "High",
      "rationale": "Dependence without domestic frontier providers"
    },
    {
      "region": "china",
      "exposure": "Medium",
      "rationale": "Domestic capability; its own displacement"
    },
    {
      "region": "indo-pacific",
      "exposure": "High",
      "rationale": "Services exporters; Japan and Korea's ageing labour markets as mitigant"
    },
    {
      "region": "south-asia",
      "exposure": "High",
      "rationale": "IT services and outsourcing exposure"
    },
    {
      "region": "gulf-middle-east",
      "exposure": "Medium",
      "rationale": "Importers with capital"
    },
    {
      "region": "africa",
      "exposure": "Medium",
      "rationale": "Lower exposure; leapfrog potential; dependence"
    },
    {
      "region": "latin-america",
      "exposure": "Medium",
      "rationale": null
    },
    {
      "region": "russia-eurasia",
      "exposure": "Medium",
      "rationale": null
    }
  ],
  "indicators": [
    {
      "name": "Entry-level vacancy and graduate employment series by occupation",
      "source": null,
      "threshold": null,
      "cadence": null
    },
    {
      "name": "Share of tasks automated in exposed sectors",
      "source": "ILO, OECD",
      "threshold": null,
      "cadence": null
    },
    {
      "name": "AI capex and model price per token",
      "source": null,
      "threshold": null,
      "cadence": null
    },
    {
      "name": "Public procurement of foreign AI services",
      "source": null,
      "threshold": null,
      "cadence": null
    },
    {
      "name": "Education application trends",
      "source": null,
      "threshold": null,
      "cadence": null
    },
    {
      "name": "Equity concentration in AI-linked stocks",
      "source": null,
      "threshold": null,
      "cadence": null
    },
    {
      "name": "Sovereign compute capacity",
      "source": null,
      "threshold": null,
      "cadence": null
    }
  ],
  "compounds": {
    "triggers": [
      {
        "id": "S03",
        "note": "market-concentration variant"
      },
      {
        "id": "S07",
        "note": "distributional conflict"
      },
      {
        "id": "S02",
        "note": "compute cut-off"
      }
    ],
    "triggered_by": [
      {
        "id": "S01",
        "note": "export controls on chips and models"
      }
    ],
    "amplified_by": [],
    "triggered_by_other": [],
    "amplifying_trends": [
      "digital sovereignty gaps",
      "demographic shifts",
      "education cost structures"
    ]
  },
  "open_questions": [
    "Whether a scenario whose GDP effect may be positive fits a stress library.",
    "Whether the market-concentration variant should be a separate scenario.",
    "How to avoid confusing the trend (diffusion) with the shock (speed and dependence)."
  ],
  "commentary": null,
  "overview": {
    "source": "ginc-desk-v0.2",
    "lenses": {
      "political": "Unions, professional bodies and three parties converge on **a tax on compute**. The compute is abroad.",
      "economic": "Output is up, yet **employment tax receipts fall** faster than the finance ministry's models allowed.",
      "social": "The graduate round that takes 40,000 **takes 9,000**. The welfare system meets the qualified, young, unemployed and online.",
      "technological": "Firms buy the work from **an API billed in a foreign currency**. Models, compute and standards sit with a few foreign providers.",
      "legal": "A provider **changes its terms of service** in a quarter and the public digital team cannot run its own systems.",
      "environmental": "**Compute power demand** and data-centre siting carry a medium load."
    },
    "regions": {
      "north-america": "**Origin**: concentration and displacement together.",
      "europe": "**Dependence** without domestic frontier providers.",
      "china": "**Domestic capability**, and its own displacement.",
      "indo-pacific": "Services exporters; **ageing labour markets** in Japan and Korea act as a mitigant.",
      "south-asia": "**IT services and outsourcing** exposure.",
      "gulf-middle-east": "**Importers with capital**.",
      "africa": "Lower exposure and leapfrog potential, with **dependence**.",
      "latin-america": "Exposed through knowledge-work displacement and **dependence on foreign models**.",
      "russia-eurasia": "Exposed through knowledge-work displacement and **dependence on foreign models**."
    }
  },
  "history": [
    {
      "edition": 2027,
      "rank": 7,
      "L2y": 3,
      "L10y": 5,
      "impact_systemic": 4,
      "impact_national": 4,
      "confidence": "low"
    }
  ],
  "changelog": [
    {
      "version": "0.2.0",
      "date": "2026-10-02",
      "change": "Entered the Library at v0.2 with GINC desk scores."
    }
  ],
  "citation": "GINC (2027). Scenario S13 Rapid AI diffusion shock, Scenario Library v0.2. scenarios.ginc.org/library/rapid-ai-diffusion-shock"
}