convez.ai
PluginSegurançaIntermediário
lexbeam-software

EU AI Governance Plugin

lexbeam-software/eu-ai-governance-plugin

Classifique seu sistema de IA conforme EU AI Act em 90 segundos

Ver no GitHub
InicialPluginScore · de 0 a 100

Plugin para classificação de conformidade com a EU AI Act, integrado ao Claude Code. Realiza avaliação de risco de sistemas de IA, análise de impacto em direitos fundamentais, inventário de sistemas e avaliações de fornecedores — com foco em requisitos DACH e obrigações de conselhos de trabalho.

Instalação

$npx convez-ai lexbeam-software/eu-ai-governance-plugin

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
1
Acompanhando
0
Saúde
25%
Último commit
há 1 dia
Versão
v1.2.2
Issues abertas
1
Contribuidores
2

Casos de uso

  • Classificação de risco de sistemas de IA segundo EU AI Act
  • Documentação de conformidade para auditorias regulatórias
  • Avaliações de fornecedores e redlines de contrato
  • Inventário e mapeamento de IA em operação
  • Análise de impacto em direitos fundamentais (DPIA-IA)

Veredito

Ferramenta especializada e sólida para governança de IA em contexto europeu, com profundo conhecimento de regulação DACH. Ideal para times legais e compliance que precisam passar da teoria regulatória para inventário operacional documentado — sem assumir responsabilidade legal, mas com rigor técnico e regulatório concreto.

README

A tradução para o português deste README ainda está na fila.

EU AI Governance Plugin

Version License EU AI Act Platform

Classify your first AI system in 90 seconds.

A Claude plugin for in-house legal, compliance, and AI governance teams. The Annex III high-risk obligations apply from 2 August 2026 in current law (a 7 May 2026 provisional Council/Parliament agreement would shift this to 2 December 2027, but is not yet adopted; current law remains authoritative until formal adoption and OJ publication). This plugin gets you from "what AI do we actually have?" to a classified inventory, vendor redlines, and auditor-ready evidence, without another SaaS login.

See it live · 20-min intro call with Werner

Install

# 1. Add the Lexbeam marketplace
claude plugin marketplace add lexbeam-software/eu-ai-governance-plugin

# 2. Install the plugin
claude plugin install eu-ai-governance@lexbeam

# 3. See where you stand
/ai-act-status

Disclaimer. This plugin is an orientation tool, not legal advice. Standardized, rule-based software run by the user in their own environment, keine Rechtsdienstleistung im Sinne des § 2 RDG. Lexbeam Software is not a Rechtsanwaltsgesellschaft and does not provide case-specific legal analysis. For concrete legal questions, consult a registered Rechtsanwältin or Rechtsanwalt. Full text: LEGAL-DISCLAIMER.md.

Built & maintained by Werner Plutat

Legal Engineer · LL.M. Business Law · 10+ years in Big 4 & DAX compliance.

I use this plugin daily on paid client mandates. That's why the regulatory content stays current the day new guidance drops, and why the DACH specifics, Works Council under BetrVG, BaFin, BSI, are actually right.

LinkedIn · lexbeam.com · Intro call

Why this exists

The EU AI Act is live. Prohibited practices are already banned. GPAI transparency obligations are in effect. High-risk Annex III obligations apply from 2 August 2026 in current law (a 7 May 2026 provisional Council/Parliament agreement would shift this to 2 December 2027 if formally adopted; not yet adopted, current law authoritative). Anthropic's legal plugin handles generic GDPR and contract work. This plugin handles what it doesn't:

  • AI Act risk classification (is your system high-risk?)
  • Provider vs. deployer obligations (what's YOUR responsibility?)
  • Fundamental rights impact assessments
  • AI inventory and documentation requirements
  • Works council considerations for AI systems (mandatory in DACH)
  • Schrems II transfer impact assessments for AI model providers

Works standalone or alongside Anthropic's legal plugin.

How It Classifies AI Risk

The plugin uses a multi-gate decision framework aligned to the EU AI Act:

flowchart TD
    A["AI System Description"] --> B{"Gate 1: Prohibited?<br/>(Article 5)"}
    B -->|Yes| C["PROHIBITED<br/>Immediate escalation"]
    B -->|No| D{"Gate 2A: Safety component<br/>of regulated product?<br/>(Annex I)"}
    D -->|Yes| E["HIGH-RISK<br/>(Annex I)"]
    D -->|No| F{"Gate 2B: High-risk<br/>use case?<br/>(Annex III)"}
    F -->|Yes| G["HIGH-RISK<br/>(Annex III)"]
    F -->|No| H{"Gate 2C: GPAI Model?<br/>(Articles 51-56)"}
    H -->|Yes| I["GPAI OBLIGATIONS<br/>apply at model level"]
    H -->|No| J{"Gate 3: Transparency<br/>duties?<br/>(Article 50)"}
    J -->|Yes| K["LIMITED RISK<br/>Transparency obligations"]
    J -->|No| L["MINIMAL RISK<br/>Voluntary codes of conduct"]

    style C fill:#dc2626,color:#fff
    style E fill:#ea580c,color:#fff
    style G fill:#ea580c,color:#fff
    style I fill:#7c3aed,color:#fff
    style K fill:#2563eb,color:#fff
    style L fill:#16a34a,color:#fff

Compliance Timeline

Key enforcement deadlines, the plugin tracks where you stand against each. Dates shown are current-law dates. A 7 May 2026 provisional Council/Parliament agreement would shift Annex III to 2 Dec 2027 and Annex I to 2 Aug 2028 if formally adopted; not yet adopted, current law authoritative.

gantt
    title EU AI Act Enforcement Timeline (current law)
    dateFormat YYYY-MM-DD
    axisFormat %b %Y

    section Already in Force
    Prohibited practices ban (Art. 5)           :done, 2025-02-02, 1d
    AI literacy obligations (Art. 4)            :done, 2025-02-02, 1d
    GPAI model obligations (Arts. 51-56)        :done, 2025-08-02, 1d

    section Upcoming
    High-risk Annex III obligations              :crit, 2026-08-02, 1d

    section Future
    High-risk Annex I obligations                :2027-08-02, 1d

Commands

CommandWhat it does
/classify-ai-riskDetermine if an AI system is high-risk under the EU AI Act. Step-by-step classification with regulatory citations.
/assess-ai-vendorReview an AI vendor/provider contract against AI Act deployer obligations, GDPR requirements, and enterprise governance standards.
/run-dpiaConduct a Data Protection Impact Assessment specifically designed for AI systems. Covers both GDPR Art. 35 and AI Act requirements.
/ai-act-statusAssess your organization's compliance posture against EU AI Act deadlines. Gap analysis with prioritized action items.
/generate-evidence-packCompile governance documentation for auditors, regulators, or internal review. Structured output ready for regulatory inspection.
/review-ai-policyReview or draft an AI governance policy against EU AI Act requirements and industry best practices.

Skills

SkillWhen it activates
AI Act ClassificationDetermining risk levels, prohibited practices, GPAI obligations
AI Vendor AssessmentEvaluating AI provider contracts and compliance posture
DPIA for AI SystemsImpact assessments combining GDPR and AI Act requirements
Governance DocumentationCreating audit trails, evidence packs, compliance records
EU Compliance (Extended)GDPR, AI Act, works council, and cross-border transfer requirements
Risk ManagementAI-specific risk assessment frameworks and controls

What makes this different from the generic legal plugin

Anthropic Legal PluginEU AI Governance Plugin
GDPR/CCPA checklistsFull EU AI Act classification engine with regulatory citations
Generic contract reviewAI vendor-specific assessment against deployer obligations
No AI Act awarenessDeadline-aware compliance (Feb 25, Aug 25, Aug 26, Aug 27)
US-centric defaultsDACH-first with German templates and works council integration
No audit trail guidanceEvidence pack generation for regulatory inspection
Manual playbook configurationPre-built EU governance playbooks, ready out of the box
No risk classificationStructured risk classification per Annex III

First 10 minutes after install

  1. Run /ai-act-status to see where your organization stands
  2. Use /classify-ai-risk for each AI system in your inventory
  3. Run /assess-ai-vendor on your AI provider contracts
  4. Generate evidence packs with /generate-evidence-pack

No configuration. No API keys. Works with what your team already has.

Language Support

  • English: Full support for all commands and outputs
  • German (Deutsch): Templates, governance documents, and compliance reports available in German. Use --lang de with any command.

Requirements

  • Claude Cowork or Claude Code
  • No external dependencies, no API keys, no infrastructure
  • Works offline with local files

Regulatory Coverage

  • EU AI Act (Regulation 2024/1689): Full classification, obligations, and timeline
  • GDPR: Extended DPA review with AI-specific considerations
  • German Works Constitution Act (BetrVG): Works council consultation requirements for AI systems
  • Schrems II / EU-US DPF: Transfer impact assessments for AI model providers
  • ISO 42001: AI management system alignment (optional)

Worked Examples

See what the plugin actually produces, realistic, redacted sample outputs:

ExampleCommandScenario
HR Resume Screening AI/classify-ai-riskClassifying an automated recruitment tool as HIGH-RISK (Annex III) with Works Council obligations
ChatGPT Enterprise Deployment/assess-ai-vendorVendor assessment with RED/YELLOW flags, contract redlines, and Schrems II analysis
Customer Churn Prediction/run-dpiaFull DPIA for an ML model processing customer behavioral data

Roadmap

VersionFocusStatus
v1.0EU AI Act classification, DPIAs, vendor assessments, evidence packs, policy reviewReleased
v1.1ISO 42001 full alignment, control mapping and certification readiness workflowsPlanned
v1.2NIS2 integration, cybersecurity obligations for AI in critical infrastructurePlanned
v1.3Multi-jurisdiction, French CNIL, Dutch AP, and Austrian DSB-specific guidancePlanned
v2.0Implementing acts and harmonised standards tracking, auto-update as EU AI Office publishes guidancePlanned

License

Apache 2.0, Fork it, extend it, use it commercially.

Built by

Lexbeam Software, AI Governance and Legal Tech for European enterprises.


Live demo · Changelog · Legal disclaimer

Extends Anthropic's knowledge-work-plugins. Not affiliated with or endorsed by Anthropic.

This plugin extends and complements Anthropic's knowledge-work-plugins. It is not affiliated with or endorsed by Anthropic.

Ver o original no GitHub

Bom para

Relacionados