Dangerous Capability Evaluations

{
  "id": "dangerous-cap-evals",
  "wikiId": "E442",
  "path": "/knowledge-base/responses/dangerous-cap-evals/",
  "filePath": "knowledge-base/responses/dangerous-cap-evals.mdx",
  "title": "Dangerous Capability Evaluations",
  "quality": 64,
  "readerImportance": 70.5,
  "researchImportance": 71.5,
  "tacticalValue": null,
  "contentFormat": "article",
  "causalLevel": null,
  "lastUpdated": "2026-01-29",
  "dateCreated": "2026-02-15",
  "summary": "Comprehensive synthesis showing dangerous capability evaluations are now standard practice (95%+ frontier models) but face critical limitations: AI capabilities double every 7 months while external safety orgs are underfunded 10,000:1 vs development, and 1-13% of models exhibit scheming behavior that could evade evaluations. Despite achieving significant adoption and identifying real deployment risks (e.g., o3 scoring 43.8% on virology tests vs 22.1% human expert average), DCEs cannot guarantee safety against sophisticated deception or emergent capabilities.",
  "description": "Systematic testing of AI models for dangerous capabilities including bioweapons assistance, cyberattack potential, autonomous self-replication, and persuasion/manipulation abilities to inform deployment decisions and safety policies.",
  "ratings": {
    "novelty": 4.5,
    "rigor": 7,
    "completeness": 7.5,
    "actionability": 6.5
  },
  "category": "responses",
  "subcategory": "alignment-evaluation",
  "clusters": [
    "ai-safety",
    "governance"
  ],
  "metrics": {
    "wordCount": 3557,
    "tableCount": 18,
    "diagramCount": 2,
    "internalLinks": 12,
    "externalLinks": 39,
    "footnoteCount": 0,
    "bulletRatio": 0.13,
    "sectionCount": 35,
    "hasOverview": true,
    "structuralScore": 15
  },
  "suggestedQuality": 100,
  "updateFrequency": 21,
  "evergreen": true,
  "wordCount": 3557,
  "unconvertedLinks": [
    {
      "text": "Autonomous task completion",
      "url": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "resourceId": "271fc5f73a8304b2",
      "resourceTitle": "Measuring AI Ability to Complete Long Tasks - METR"
    },
    {
      "text": "GPT-5 Evaluation",
      "url": "https://evaluations.metr.org/gpt-5-report/",
      "resourceId": "7457262d461e2206",
      "resourceTitle": "Details about METR’s evaluation of OpenAI GPT-5"
    },
    {
      "text": "Frontier AI Trends",
      "url": "https://www.aisi.gov.uk/frontier-ai-trends-report",
      "resourceId": "7042c7f8de04ccb1",
      "resourceTitle": "AISI Frontier AI Trends"
    },
    {
      "text": "In-context scheming",
      "url": "https://www.apolloresearch.ai/research/scheming-reasoning-evaluations",
      "resourceId": "91737bf431000298",
      "resourceTitle": "Frontier Models are Capable of In-Context Scheming"
    },
    {
      "text": "Anti-scheming training",
      "url": "https://www.apolloresearch.ai/blog/more-capable-models-are-better-at-in-context-scheming/",
      "resourceId": "80c6d6eca17dc925",
      "resourceTitle": "More capable models scheme at higher rates"
    },
    {
      "text": "Dangerous Capabilities",
      "url": "https://arxiv.org/abs/2403.13793",
      "resourceId": "daec8c61ea79836b",
      "resourceTitle": "Dangerous Capability Evaluations"
    },
    {
      "text": "METR",
      "url": "https://metr.org/",
      "resourceId": "45370a5153534152",
      "resourceTitle": "METR: Model Evaluation and Threat Research"
    },
    {
      "text": "UK AI Security Institute",
      "url": "https://www.aisi.gov.uk/",
      "resourceId": "fdf68a8f30f57dee",
      "resourceTitle": "UK AI Safety Institute (AISI)"
    },
    {
      "text": "Frontier AI Trends Report",
      "url": "https://www.aisi.gov.uk/frontier-ai-trends-report",
      "resourceId": "7042c7f8de04ccb1",
      "resourceTitle": "AISI Frontier AI Trends"
    },
    {
      "text": "DeepMind's March 2024 research",
      "url": "https://arxiv.org/abs/2403.13793",
      "resourceId": "daec8c61ea79836b",
      "resourceTitle": "Dangerous Capability Evaluations"
    },
    {
      "text": "Future of Life Institute's 2025 AI Safety Index",
      "url": "https://futureoflife.org/ai-safety-index-summer-2025/",
      "resourceId": "df46edd6fa2078d1",
      "resourceTitle": "FLI AI Safety Index Summer 2025"
    },
    {
      "text": "Responsible Scaling Policy",
      "url": "https://www.anthropic.com/responsible-scaling-policy",
      "resourceId": "afe1e125f3ba3f14",
      "resourceTitle": "Responsible Scaling Policy"
    },
    {
      "text": "Preparedness Framework",
      "url": "https://openai.com/index/updating-our-preparedness-framework/",
      "resourceId": "ded0b05862511312",
      "resourceTitle": "Preparedness Framework"
    },
    {
      "text": "Apollo Research's December 2024 study",
      "url": "https://www.apolloresearch.ai/research/scheming-reasoning-evaluations",
      "resourceId": "91737bf431000298",
      "resourceTitle": "Frontier Models are Capable of In-Context Scheming"
    },
    {
      "text": "follow-up anti-scheming training research",
      "url": "https://openai.com/index/detecting-and-reducing-scheming-in-ai-models/",
      "resourceId": "b3f335edccfc5333",
      "resourceTitle": "OpenAI Preparedness Framework"
    },
    {
      "text": "International AI Safety Report's October 2025 update",
      "url": "https://internationalaisafetyreport.org/publication/first-key-update-capabilities-and-risk-implications",
      "resourceId": "6acf3be7a03c2328",
      "resourceTitle": "International AI Safety Report (October 2025)"
    },
    {
      "text": "Evaluating Frontier Models for Dangerous Capabilities",
      "url": "https://arxiv.org/abs/2403.13793",
      "resourceId": "daec8c61ea79836b",
      "resourceTitle": "Dangerous Capability Evaluations"
    },
    {
      "text": "Measuring AI Ability to Complete Long Tasks",
      "url": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "resourceId": "271fc5f73a8304b2",
      "resourceTitle": "Measuring AI Ability to Complete Long Tasks - METR"
    },
    {
      "text": "Frontier AI Trends Report",
      "url": "https://www.aisi.gov.uk/frontier-ai-trends-report",
      "resourceId": "7042c7f8de04ccb1",
      "resourceTitle": "AISI Frontier AI Trends"
    },
    {
      "text": "Detecting and Reducing Scheming",
      "url": "https://openai.com/index/detecting-and-reducing-scheming-in-ai-models/",
      "resourceId": "b3f335edccfc5333",
      "resourceTitle": "OpenAI Preparedness Framework"
    },
    {
      "text": "First Key Update: Capabilities and Risk Implications",
      "url": "https://internationalaisafetyreport.org/publication/first-key-update-capabilities-and-risk-implications",
      "resourceId": "6acf3be7a03c2328",
      "resourceTitle": "International AI Safety Report (October 2025)"
    },
    {
      "text": "AI Safety Index Summer 2025",
      "url": "https://futureoflife.org/ai-safety-index-summer-2025/",
      "resourceId": "df46edd6fa2078d1",
      "resourceTitle": "FLI AI Safety Index Summer 2025"
    },
    {
      "text": "Our 2025 Year in Review",
      "url": "https://www.aisi.gov.uk/blog/our-2025-year-in-review",
      "resourceId": "3dec5f974c5da5ec",
      "resourceTitle": "Our 2025 Year in Review"
    },
    {
      "text": "Advanced AI Evaluations May Update",
      "url": "https://www.aisi.gov.uk/blog/advanced-ai-evaluations-may-update",
      "resourceId": "4e56cdf6b04b126b",
      "resourceTitle": "UK AI Safety Institute renamed to AI Security Institute"
    },
    {
      "text": "Responsible Scaling Policy v2.2",
      "url": "https://www.anthropic.com/responsible-scaling-policy",
      "resourceId": "afe1e125f3ba3f14",
      "resourceTitle": "Responsible Scaling Policy"
    },
    {
      "text": "Preparedness Framework v2",
      "url": "https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf",
      "resourceId": "ec5d8e7d6a1b2c7c",
      "resourceTitle": "OpenAI: Preparedness Framework Version 2"
    },
    {
      "text": "Common Elements of Frontier AI Safety Policies",
      "url": "https://metr.org/assets/common-elements-nov-2024.pdf",
      "resourceId": "fdacc4a389a5f5f7",
      "resourceTitle": "[PDF] Common Elements of Frontier AI Safety Policies - METR"
    },
    {
      "text": "METR",
      "url": "https://metr.org/",
      "resourceId": "45370a5153534152",
      "resourceTitle": "METR: Model Evaluation and Threat Research"
    },
    {
      "text": "Apollo Research",
      "url": "https://www.apolloresearch.ai/",
      "resourceId": "329d8c2e2532be3d",
      "resourceTitle": "Apollo Research - AI Safety Evaluation Organization"
    },
    {
      "text": "UK AI Security Institute",
      "url": "https://www.aisi.gov.uk/",
      "resourceId": "fdf68a8f30f57dee",
      "resourceTitle": "UK AI Safety Institute (AISI)"
    },
    {
      "text": "US AI Safety Institute (NIST)",
      "url": "https://www.nist.gov/aisi",
      "resourceId": "84e0da6d5092e27d",
      "resourceTitle": "Center for AI Standards and Innovation (CAISI)"
    },
    {
      "text": "SecureBio",
      "url": "https://securebio.org/",
      "resourceId": "81e8568b008e4245",
      "resourceTitle": "SecureBio organization"
    }
  ],
  "unconvertedLinkCount": 32,
  "convertedLinkCount": 0,
  "backlinkCount": 5,
  "hallucinationRisk": {
    "level": "low",
    "score": 30,
    "factors": [
      "no-citations",
      "high-rigor",
      "conceptual-content"
    ]
  },
  "entityType": "approach",
  "redundancy": {
    "maxSimilarity": 23,
    "similarPages": [
      {
        "id": "capability-elicitation",
        "title": "Capability Elicitation",
        "path": "/knowledge-base/responses/capability-elicitation/",
        "similarity": 23
      },
      {
        "id": "model-auditing",
        "title": "Third-Party Model Auditing",
        "path": "/knowledge-base/responses/model-auditing/",
        "similarity": 23
      },
      {
        "id": "evals",
        "title": "Evals & Red-teaming",
        "path": "/knowledge-base/responses/evals/",
        "similarity": 22
      },
      {
        "id": "alignment-evals",
        "title": "Alignment Evaluations",
        "path": "/knowledge-base/responses/alignment-evals/",
        "similarity": 20
      },
      {
        "id": "capability-threshold-model",
        "title": "Capability Threshold Model",
        "path": "/knowledge-base/models/capability-threshold-model/",
        "similarity": 18
      }
    ]
  },
  "coverage": {
    "passing": 8,
    "total": 13,
    "targets": {
      "tables": 14,
      "diagrams": 1,
      "internalLinks": 28,
      "externalLinks": 18,
      "footnotes": 11,
      "references": 11
    },
    "actuals": {
      "tables": 18,
      "diagrams": 2,
      "internalLinks": 12,
      "externalLinks": 39,
      "footnotes": 0,
      "references": 20,
      "quotesWithQuotes": 0,
      "quotesTotal": 0,
      "accuracyChecked": 0,
      "accuracyTotal": 0
    },
    "items": {
      "summary": "green",
      "schedule": "green",
      "entity": "green",
      "editHistory": "red",
      "overview": "green",
      "tables": "green",
      "diagrams": "green",
      "internalLinks": "amber",
      "externalLinks": "green",
      "footnotes": "red",
      "references": "green",
      "quotes": "red",
      "accuracy": "red"
    },
    "ratingsString": "N:4.5 R:7 A:6.5 C:7.5"
  },
  "readerRank": 162,
  "researchRank": 137,
  "recommendedScore": 167.41
}

No external links