{
  "schema": "https://policywindow.org/schemas/notebook-bundle.v1.json",
  "manifest": {
    "bundleVersion": 1,
    "generatedAt": "2026-06-01",
    "instrument": {
      "shortCode": "EU-AIA-2024",
      "name": "EU AI Act",
      "jurisdiction": "EU",
      "kind": "binding_regulation",
      "status": "in_force",
      "adoptedDate": "2024-07-12",
      "effectiveDate": "2024-08-01",
      "sourceUrl": "https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32024R1689",
      "sourceCitation": "Regulation (EU) 2024/1689"
    },
    "canonicalUrl": "https://policywindow.org/wiki/eu-ai-act?asOf=2026-06-01",
    "liveCanonicalUrl": "https://policywindow.org/wiki/eu-ai-act",
    "asOf": "2026-06-01",
    "license": {
      "articleContent": "CC BY 4.0",
      "citationGraph": "CC0 1.0",
      "bundleEnvelope": "CC0 1.0"
    },
    "attribution": "Policy Window (https://policywindow.org/wiki/eu-ai-act?asOf=2026-06-01). Article content CC BY 4.0; citation graph CC0 1.0. Catalog reflects editorial-cadence updates and the methodology §11 disclosed limits."
  },
  "files": {
    "instrument.md": "# EU AI Act\n\n**Jurisdiction:** EU  \n**Kind:** binding regulation  \n**Status:** in force  \n**Adopted:** 2024-07-12  \n**Effective:** 2024-08-01  \n**Last editorial review:** 2026-05-24  \n\n## Scope and obligations\n\nRisk-based framework. Prohibited practices (Art. 5) effective 2 February 2025; general-purpose AI obligations (Arts. 51-55) 2 August 2025; high-risk system obligations (Title III) 2 August 2026. Staggered 6/12/24-month application timeline from 1 August 2024 entry-into-force per Regulation (EU) 2024/1689 Art. 113.\n\n## Decision-support header\n\n**Policy question:** What obligations does the EU AI Act impose on a deployer of a high-risk AI system, and when do they take effect?  \n**Current answer:** Deployers of Annex III high-risk systems must conduct a Fundamental Rights Impact Assessment (Art. 27), monitor system operation (Art. 26(5)), retain logs (Art. 26(6)), inform affected workers (Art. 26(7)), and ensure human oversight (Art. 26(2)). Obligations apply from 2 August 2026; the prohibited-practices regime (Art. 5) is already in force since February 2025, and the general-purpose AI obligations (Art. 51-55) since August 2025.  \n**Confidence:** high  \n\n## Coverage matrix (this instrument × catalogued topics)\n\n| Topic | Coverage | Citation | Power-asymmetry analysis |\n|-------|----------|----------|--------------------------|\n| Foundation Models / GPAI | governs | Arts. 51-55 (general-purpose AI + systemic risk) | — |\n| Biometric Identification | governs | Art. 5(1)(h) prohibition + Art. 26(10) post-hoc rules | Art. 5(1)(h) prohibits real-time remote biometric identification in publicly-accessible spaces, but Art. 5(1)(h)(i)-(iii) carves out three law-enforcement use cases (targeted victim search, prevention of imminent threats incl. terrorism, and tracking serious-crime suspects under Annex II). Art. 26(10) then permits post-hoc RBI subject only to ex post judicial authorisation 'without undue delay, at the latest within 24 hours' — meaning the operational default for state security actors is permitted-with-procedural-overlay, not prohibited. Carve-outs swallow the headline rule for the highest-stakes deployment context. |\n| Deepfakes / Synthetic Content | governs | Art. 50(4) (disclosure obligation for deep fakes) | — |\n| AI in Employment | governs | Annex III §4 (high-risk: employment management) | — |\n| AI in Healthcare | governs | Annex III §5(a) (high-risk: essential services) + MDR overlap | — |\n| AI in Criminal Justice | governs | Annex III §6 (high-risk: law enforcement) | — |\n| AI in Education | governs | Annex III §3 (high-risk: educational access) | — |\n| Compute-Threshold Reporting | governs | Art. 52 + Annex XIII (10²⁵ FLOP presumption) | — |\n| Transparency Obligations | governs | Arts. 13, 50 (transparency obligations) | Art. 50 transparency obligations on emotion-recognition / biometric-categorisation / deepfake systems do not apply to military, defence, or national-security deployments — Art. 2(3) excludes these entirely from the Regulation's scope. Within scope, Art. 50(2) further permits omission of synthetic-content disclosure where use is 'authorised by law to detect, prevent, investigate or prosecute criminal offences'. The disclosure floor therefore reaches private-sector deployers but not the most surveillance-heavy state contexts. |\n| Individual Redress | governs | Art. 85 (right to lodge complaints) | — |\n| Training-Data Rights | implicit | Recital 105; CDSM Directive provides primary copyright framework | — |\n| Sovereign AI Doctrine | silent | No explicit sovereign-AI doctrine | — |\n| Catastrophic & Existential Risk | implicit | Art. 51 + Recital 32 — systemic risk overlaps with but does not fully cover catastrophic-risk framing | — |\n| Technological Sovereignty | implicit | Recitals 1-5 + EU competence framing; AI Office establishes EU capacity | — |\n| Development-Rights Framings | silent | EU framework is rights-based but rooted in EU-charter rights, not development-rights doctrine | — |\n| Agentic AI Governance | implicit | Arts. 26-29 deployer obligations apply to agent operators; Arts. 51-55 GPAI obligations capture the underlying model | — |\n| Open-Weight Frontier Release | governs | Art. 53(2) + Recital 102/104 — explicit open-source GPAI exemption (with caveats for systemic-risk models) | — |\n| Synthetic Content Provenance | governs | Art. 50(2) — provider machine-readable marking obligation; Art. 50(4) — deployer disclosure for deep fakes (distinct from the `deepfakes` topic which focuses on misuse-harms) | — |\n| AI in Elections | implicit | Art. 5 prohibitions (subliminal manipulation) + Annex III §8 (democratic processes / elections high-risk) | — |\n| Compute + Model-Weight Export Controls | silent | EU AIA does not address compute / weight export controls; lives in dual-use Regulation (EU) 2021/821 | — |\n| Environmental Impact of AI Training | implicit | Art. 95 voluntary codes of conduct include environmental sustainability; Recital 142 references energy efficiency reporting for GPAI | — |\n| National Security Carveouts in AI Regulation | governs | Art. 2(3) explicitly excludes AI systems used exclusively for military, defence, or national-security purposes | — |\n| AI-Driven Worker Displacement | silent | EU AIA focuses on AI-in-employment-decisions (Annex III §4); displacement-as-cause not separately addressed | — |\n\n## Primary source\n\n- Source URL: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32024R1689\n- Source citation: Regulation (EU) 2024/1689\n\n## Canonical Policy Window URL\n\n- Live: https://policywindow.org/wiki/eu-ai-act\n- Snapshot-pinned: https://policywindow.org/wiki/eu-ai-act?asOf=2026-06-01\n\n## Attribution (reproduce verbatim under any answer)\n\nPolicy Window (https://policywindow.org/wiki/eu-ai-act?asOf=2026-06-01). Article content CC BY 4.0; citation graph CC0 1.0. Catalog reflects editorial-cadence updates and the methodology §11 disclosed limits.\n",
    "primary-source.url": "https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32024R1689\n",
    "citations.bib": "@misc{policywindow_eu_ai_act_20260601,\n  author       = {{Policy Window Contributors}},\n  title        = {EU AI Act},\n  year         = {2026},\n  month        = {jun},\n  howpublished = {Policy Window AI Governance Catalog},\n  url          = {https://policywindow.org/wiki/eu-ai-act?asOf=2026-06-01},\n  urldate      = {2026-06-01},\n  note         = {Snapshot 2026-06-01; CC0 1.0 Universal for citation graph, CC BY 4.0 for article content},\n  keywords     = {AI governance, EU, binding regulation}\n}\n",
    "README.md": "# Policy Window — NotebookLM source bundle\n\n**Instrument:** EU AI Act (EU)  \n**Snapshot:** 2026-06-01  \n**Canonical URL:** https://policywindow.org/wiki/eu-ai-act?asOf=2026-06-01\n\n## What this bundle contains\n\nEach entry under `files` is the content of a file you can create locally and attach as a NotebookLM source (or a Claude Project knowledge file, or a Custom GPT knowledge file). The JSON envelope is the transport format; the individual files are the ingest units.\n\n- `instrument.md` — markdown summary of the instrument with the catalog's cross-topic coverage table\n- `primary-source.url` — official upstream URL\n- `citations.bib` — single BibTeX entry for citation managers\n- `README.md` — this file\n\n## How to create files locally from this JSON\n\n```sh\n# Fetch the bundle\ncurl -o bundle.json 'https://policywindow.org/wiki/eu-ai-act/notebook-bundle?asOf=2026-06-01'\n\n# Extract each file (requires jq)\nmkdir eu-ai-act-notebook-bundle && cd eu-ai-act-notebook-bundle\njq -r '.files[\"instrument.md\"]' ../bundle.json > instrument.md\njq -r '.files[\"primary-source.url\"]' ../bundle.json > primary-source.url\njq -r '.files[\"citations.bib\"]' ../bundle.json > citations.bib\njq -r '.files[\"README.md\"]' ../bundle.json > README.md\n```\n\n## How to attach to NotebookLM\n\n1. Visit https://notebooklm.google.com and create a new notebook.\n2. Click \"+ Add source\" for each of the 4 extracted files.\n3. Paste the conversation-focus statement from https://policywindow.org/wiki/templates/notebooklm into the Customise panel.\n4. Ask questions — NotebookLM surfaces citation-by-passage pop-outs for every claim.\n\n## Attribution requirement (reproduce verbatim under any agent answer)\n\nPolicy Window (https://policywindow.org/wiki/eu-ai-act?asOf=2026-06-01). Article content CC BY 4.0; citation graph CC0 1.0. Catalog reflects editorial-cadence updates and the methodology §11 disclosed limits.\n\n## Snapshot pinning\n\nThis bundle is pinned to the 2026-06-01 snapshot. To get the live version, omit the asOf query string: https://policywindow.org/wiki/eu-ai-act/notebook-bundle.\n\nBundle license: CC0 1.0 Universal (the JSON envelope itself is freely re-distributable; the article content under CC BY 4.0 + the citation graph under CC0 1.0 — see /wiki/funding for the licensing matrix).\n"
  }
}