borghei/Claude-Skills / ra-qm-team/eu-ai-act-specialist
eu-ai-act-specialist
VerifiedEU AI Act (Regulation EU 2024/1689) compliance specialist. Use for AI system risk classification, provider/deployer obligations, GPAI model compliance, conformity assessments, bias and fairness testing, and AI governance programs.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 28.5 KB |
| references/ai-act-classification-guide.md | markdown | 28.2 KB |
| references/ai-governance-framework.md | markdown | 32.1 KB |
| references/ai-technical-documentation-templates.md | markdown | 70.7 KB |
| scripts/ai_bias_detector.py | script | 31.9 KB |
| scripts/ai_compliance_checker.py | script | 55.3 KB |
| scripts/ai_risk_classifier.py | script | 42.3 KB |
scripts/ai_compliance_checker.py
View on GitHub#!/usr/bin/env python3
"""
EU AI Act Compliance Checker
Validates AI system compliance status against all provider and deployer
obligations under Regulation (EU) 2024/1689. Generates gap analysis with
remediation steps and scores overall compliance readiness (0-100).
Usage:
python ai_compliance_checker.py --input compliance_status.json
python ai_compliance_checker.py --input compliance_status.json --json
python ai_compliance_checker.py --input compliance_status.json --role deployer
python ai_compliance_checker.py --input compliance_status.json --output report.json
Input JSON schema:
{
"system_name": "AI System Name",
"risk_level": "HIGH",
"role": "provider",
"is_gpai": false,
"is_systemic_risk": false,
"is_public_body": false,
"risk_management": {
"process_established": true,
"risks_identified": true,
"risks_evaluated": true,
"measures_adopted": true,
"residual_risk_acceptable": true,
"testing_performed": true,
"lifecycle_coverage": true
},
"data_governance": {
"design_choices_documented": true,
"preparation_documented": true,
"assumptions_documented": true,
"data_sufficient": true,
"data_representative": true,
"bias_examined": true,
"bias_mitigated": true,
"special_categories_handled": true
},
"technical_documentation": {
"system_description": true,
"development_process": true,
"monitoring_description": true,
"risk_management_docs": true,
"data_governance_docs": true,
"performance_metrics": true,
"training_data_info": true,
"lifecycle_changes": true,
"kept_up_to_date": true
},
"record_keeping": {
"automatic_logging": true,
"usage_period_logged": true,
"reference_db_logged": true,
"input_matches_logged": true,
"verification_persons_logged": true,
"retention_period_defined": true
},
"transparency": {
"instructions_provided": true,
"provider_identity": true,
"capabilities_limitations": true,
"intended_purpose_stated": true,
"accuracy_levels_declared": true,
"risk_circumstances_documented": true,
"group_performance_documented": true,
"input_specs_provided": true,
"oversight_measures_described": true,
"resource_needs_stated": true,
"lifetime_maintenance_described": true,
"logging_mechanism_described": true
},
"human_oversight": {
"oversight_level_defined": true,
"understand_capacities": true,
"interpret_output": true,
"can_disregard_output": true,
"intervention_mechanism": true,
"automation_bias_safeguards": true
},
"accuracy_robustness_cybersecurity": {
"accuracy_levels_achieved": true,
"robustness_ensured": true,
"redundancy_failsafe": true,
"data_poisoning_protection": true,
"adversarial_protection": true,
"model_manipulation_protection": true,
"confidentiality_protection": true,
"security_proportionate": true
},
"quality_management": {
"compliance_strategy": true,
"design_verification": true,
"development_processes": true,
"test_procedures": true,
"standards_documented": true,
"data_management_systems": true,
"risk_management_integrated": true,
"post_market_monitoring": true,
"incident_reporting_procedures": true,
"authority_communication": true,
"record_keeping_procedures": true,
"resource_management": true,
"accountability_framework": true
},
"conformity_assessment": {
"path_determined": true,
"documentation_compiled": true,
"assessment_completed": true,
"declaration_signed": true
},
"ce_marking": {
"marking_affixed": true,
"notified_body_number": true,
"precedes_market_placement": true
},
"eu_database": {
"provider_registered": true,
"system_description_entered": true,
"conformity_status_entered": true,
"member_states_entered": true,
"restrictions_documented": true
},
"post_market_monitoring": {
"system_established": true,
"data_collection_defined": true,
"risk_management_updated": true,
"plan_in_documentation": true,
"incident_reporting_defined": true,
"interaction_analysis": true
},
"deployer": {
"use_per_instructions": true,
"human_oversight_assigned": true,
"input_data_relevant": true,
"monitoring_active": true,
"logs_retained": true,
"workers_informed": true,
"dpia_completed": true,
"fria_completed": true,
"incidents_reported": true
},
"gpai": {
"technical_docs": true,
"downstream_info": true,
"copyright_policy": true,
"training_summary_public": true,
"eu_representative": true
},
"gpai_systemic": {
"model_evaluation": true,
"red_teaming": true,
"risk_assessment": true,
"incident_tracking": true,
"cybersecurity": true,
"energy_reporting": true
}
}
"""
import argparse
import json
import sys
from datetime import datetime
from typing import Any, Dict, List, Optional, Tuple
# ---------------------------------------------------------------------------
# Obligation definitions with weights and remediation guidance
# ---------------------------------------------------------------------------
PROVIDER_OBLIGATION_SECTIONS = {
"risk_management": {
"article": "Art. 9",
"title": "Risk Management System",
"weight": 12,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"process_established": {
"label": "Continuous iterative risk management process established",
"remediation": "Establish a documented risk management process that operates throughout the AI system lifecycle. Define risk identification, evaluation, and control procedures. Assign a risk management owner.",
"priority": "HIGH",
},
"risks_identified": {
"label": "Known and foreseeable risks identified and analysed",
"remediation": "Conduct a comprehensive risk identification workshop covering health, safety, and fundamental rights risks. Use structured techniques (FMEA, HAZOP, fault tree analysis).",
"priority": "HIGH",
},
"risks_evaluated": {
"label": "Risks estimated and evaluated for intended use and foreseeable misuse",
"remediation": "Evaluate each risk for likelihood and severity under both intended use and reasonably foreseeable misuse scenarios. Document evaluation criteria and results.",
"priority": "HIGH",
},
"measures_adopted": {
"label": "Suitable risk management measures adopted",
"remediation": "Define and implement risk control measures for each identified risk. Prioritize elimination, then reduction, then protective measures. Document rationale for each measure.",
"priority": "HIGH",
},
"residual_risk_acceptable": {
"label": "Residual risk determined acceptable",
"remediation": "Assess residual risk after control measures. Document acceptability criteria, the risk-benefit analysis, and the basis for determining residual risks are acceptable.",
"priority": "HIGH",
},
"testing_performed": {
"label": "System tested against defined metrics and thresholds",
"remediation": "Define quantitative performance metrics and acceptable thresholds. Test the system against these before deployment and periodically throughout lifecycle.",
"priority": "MEDIUM",
},
"lifecycle_coverage": {
"label": "Risk management covers full system lifecycle",
"remediation": "Ensure the risk management process extends from design through development, deployment, operation, and decommissioning. Include post-market monitoring feedback loops.",
"priority": "MEDIUM",
},
},
},
"data_governance": {
"article": "Art. 10",
"title": "Data and Data Governance",
"weight": 11,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"design_choices_documented": {
"label": "Data collection design choices documented",
"remediation": "Document the rationale for data collection methods, sources selected, volume targets, and labelling strategies. Include what data was excluded and why.",
"priority": "HIGH",
},
"preparation_documented": {
"label": "Data preparation processes documented (annotation, labelling, cleaning)",
"remediation": "Create detailed documentation of all data preparation steps: annotation guidelines, labelling protocols, cleaning rules, enrichment processes, and quality checks applied.",
"priority": "HIGH",
},
"assumptions_documented": {
"label": "Assumptions about data representation documented",
"remediation": "Document what the data is meant to measure and represent. State assumptions about the relationship between training data and the real-world population/use case.",
"priority": "MEDIUM",
},
"data_sufficient": {
"label": "Data volume sufficient for intended purpose",
"remediation": "Assess whether the quantity of training, validation, and testing data is adequate. Document the analysis and any limitations due to data scarcity.",
"priority": "MEDIUM",
},
"data_representative": {
"label": "Data representative of target population",
"remediation": "Analyse whether the dataset demographics and characteristics match the intended deployment population. Address any underrepresentation gaps.",
"priority": "HIGH",
},
"bias_examined": {
"label": "Datasets examined for possible biases",
"remediation": "Conduct systematic bias examination across protected attributes (age, gender, ethnicity, disability). Use statistical tests and the ai_bias_detector.py tool.",
"priority": "HIGH",
},
"bias_mitigated": {
"label": "Bias mitigation measures implemented",
"remediation": "Implement appropriate bias mitigation strategies: data augmentation, re-sampling, re-weighting, or algorithmic fairness constraints. Document measures and their effectiveness.",
"priority": "HIGH",
},
"special_categories_handled": {
"label": "Special category data processing handled appropriately",
"remediation": "If processing special category data (Art. 9 GDPR) for bias detection/correction, ensure strict necessity is documented and appropriate safeguards (encryption, access controls, purpose limitation) are in place.",
"priority": "MEDIUM",
},
},
},
"technical_documentation": {
"article": "Art. 11",
"title": "Technical Documentation",
"weight": 10,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"system_description": {
"label": "General system description and intended purpose documented",
"remediation": "Create a comprehensive system description covering: model architecture, input/output modalities, intended purpose, target users, affected persons, and deployment context.",
"priority": "HIGH",
},
"development_process": {
"label": "Development process and system elements described",
"remediation": "Document the system design, development methodology, component architecture, training procedures, and validation approach. Include version history.",
"priority": "HIGH",
},
"monitoring_description": {
"label": "Monitoring, functioning, and control mechanisms described",
"remediation": "Document how the system is monitored in operation, how it functions (decision logic), and what control mechanisms exist for operators.",
"priority": "MEDIUM",
},
"risk_management_docs": {
"label": "Risk management documentation included",
"remediation": "Include or reference the complete risk management system documentation (Art. 9) within the technical file.",
"priority": "HIGH",
},
"data_governance_docs": {
"label": "Data governance documentation included",
"remediation": "Include or reference the complete data governance documentation (Art. 10) within the technical file.",
"priority": "HIGH",
},
"performance_metrics": {
"label": "Performance metrics documented, including for specific groups",
"remediation": "Document all performance metrics (accuracy, precision, recall, F1, etc.) with breakdowns by relevant demographic groups and deployment contexts.",
"priority": "HIGH",
},
"training_data_info": {
"label": "Training data information included or referenced",
"remediation": "Include a description of training datasets or reference where this information is maintained. Cover data sources, volumes, characteristics, and provenance.",
"priority": "MEDIUM",
},
"lifecycle_changes": {
"label": "Lifecycle changes documented",
"remediation": "Establish a change log documenting all significant changes to the system throughout its lifecycle, including model updates, data refreshes, and configuration changes.",
"priority": "MEDIUM",
},
"kept_up_to_date": {
"label": "Technical documentation kept up to date",
"remediation": "Implement a process to review and update technical documentation at defined intervals and after any significant change to the system.",
"priority": "MEDIUM",
},
},
},
"record_keeping": {
"article": "Art. 12",
"title": "Record-Keeping / Automatic Logging",
"weight": 7,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"automatic_logging": {
"label": "Automatic event logging implemented",
"remediation": "Implement automated logging infrastructure that records system events, decisions, inputs, and outputs throughout the system's operation.",
"priority": "HIGH",
},
"usage_period_logged": {
"label": "Usage period (start/end) logged for each use",
"remediation": "Ensure logs capture the start and end timestamp of each usage session or inference request.",
"priority": "MEDIUM",
},
"reference_db_logged": {
"label": "Reference database for input checking logged",
"remediation": "Log the version or state of any reference database used for input validation or matching during each operation.",
"priority": "MEDIUM",
},
"input_matches_logged": {
"label": "Input data matches logged",
"remediation": "Log cases where input data matched entries in a reference database, including the specific match results.",
"priority": "MEDIUM",
},
"verification_persons_logged": {
"label": "Verification personnel identified in logs",
"remediation": "Log the identity of natural persons involved in reviewing or verifying system outputs.",
"priority": "MEDIUM",
},
"retention_period_defined": {
"label": "Log retention period defined and implemented",
"remediation": "Define a log retention period appropriate to the system's intended purpose. Ensure minimum compliance with EU/national law requirements. Implement automated retention management.",
"priority": "HIGH",
},
},
},
"transparency": {
"article": "Art. 13",
"title": "Transparency and Information to Deployers",
"weight": 9,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"instructions_provided": {
"label": "Instructions for use provided with the system",
"remediation": "Create comprehensive instructions for use covering all items required by Art. 13. Package with the system delivery.",
"priority": "HIGH",
},
"provider_identity": {
"label": "Provider identity and contact details included",
"remediation": "Include the provider's legal name, registered address, and contact information in the instructions for use.",
"priority": "LOW",
},
"capabilities_limitations": {
"label": "System characteristics, capabilities, and limitations described",
"remediation": "Document what the system can and cannot do. Include known edge cases, failure modes, and conditions where performance degrades.",
"priority": "HIGH",
},
"intended_purpose_stated": {
"label": "Intended purpose and foreseeable misuse stated",
"remediation": "Clearly state the intended purpose and explicitly describe foreseeable misuse scenarios with warnings against them.",
"priority": "HIGH",
},
"accuracy_levels_declared": {
"label": "Accuracy, robustness, and cybersecurity levels declared",
"remediation": "Quantify and declare the system's tested accuracy levels, robustness characteristics, and cybersecurity posture in the instructions for use.",
"priority": "HIGH",
},
"risk_circumstances_documented": {
"label": "Known risk circumstances documented",
"remediation": "Document all known or foreseeable circumstances that may create risks to health, safety, or fundamental rights.",
"priority": "MEDIUM",
},
"group_performance_documented": {
"label": "Performance for specific groups documented",
"remediation": "Document performance metrics disaggregated by relevant demographic groups (age, gender, ethnicity, etc.) the system will affect.",
"priority": "HIGH",
},
"input_specs_provided": {
"label": "Input data specifications provided",
"remediation": "Document the specifications for input data: format, quality requirements, constraints, and any pre-processing requirements.",
"priority": "MEDIUM",
},
"oversight_measures_described": {
"label": "Human oversight measures described",
"remediation": "Describe the human oversight mechanisms available and how deployers should implement effective oversight.",
"priority": "HIGH",
},
"resource_needs_stated": {
"label": "Computational and hardware resource needs stated",
"remediation": "Document the computational and hardware requirements for operating the system, including minimum and recommended specifications.",
"priority": "LOW",
},
"lifetime_maintenance_described": {
"label": "Expected lifetime and maintenance described",
"remediation": "State the expected operational lifetime and describe maintenance, update, and support procedures.",
"priority": "LOW",
},
"logging_mechanism_described": {
"label": "Logging mechanism described in instructions",
"remediation": "Describe the automatic logging mechanism (Art. 12) in the instructions for use so deployers understand what is logged and how to access logs.",
"priority": "MEDIUM",
},
},
},
"human_oversight": {
"article": "Art. 14",
"title": "Human Oversight",
"weight": 10,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"oversight_level_defined": {
"label": "Oversight level defined (in-the-loop / on-the-loop / in-command)",
"remediation": "Define the appropriate oversight level based on the risk profile and decision impact. Document the rationale for the chosen level.",
"priority": "HIGH",
},
"understand_capacities": {
"label": "Overseers can understand system capacities and limitations",
"remediation": "Provide training materials and documentation that enable oversight personnel to fully understand the system's capabilities, limitations, and typical failure modes.",
"priority": "HIGH",
},
"interpret_output": {
"label": "Overseers can correctly interpret system output",
"remediation": "Design output interfaces that are interpretable. Provide guidance on how to read and evaluate system outputs, including confidence scores and uncertainty indicators.",
"priority": "HIGH",
},
"can_disregard_output": {
"label": "Overseers can disregard or reverse system output",
"remediation": "Implement clear mechanisms allowing oversight personnel to override, disregard, or reverse any system decision or recommendation.",
"priority": "HIGH",
},
"intervention_mechanism": {
"label": "Intervention/interruption mechanism (stop button) implemented",
"remediation": "Implement a readily accessible mechanism to interrupt or stop the system's operation. Test that it functions reliably under all operating conditions.",
"priority": "HIGH",
},
"automation_bias_safeguards": {
"label": "Automation bias safeguards implemented",
"remediation": "Implement safeguards against over-reliance on AI output: mandatory review steps, randomized verification, confidence thresholds requiring human judgment, and regular calibration exercises.",
"priority": "HIGH",
},
},
},
"accuracy_robustness_cybersecurity": {
"article": "Art. 15",
"title": "Accuracy, Robustness, and Cybersecurity",
"weight": 9,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"accuracy_levels_achieved": {
"label": "Appropriate accuracy levels achieved and declared",
"remediation": "Establish accuracy benchmarks appropriate for the intended purpose. Test against these benchmarks and declare results in instructions for use.",
"priority": "HIGH",
},
"robustness_ensured": {
"label": "Robustness against errors/faults ensured",
"remediation": "Test system resilience to input errors, noisy data, missing values, and unexpected inputs. Document robustness characteristics.",
"priority": "HIGH",
},
"redundancy_failsafe": {
"label": "Redundancy and fail-safe measures in place",
"remediation": "Implement redundancy for critical components and fail-safe mechanisms that default to a safe state when errors occur.",
"priority": "MEDIUM",
},
"data_poisoning_protection": {
"label": "Protection against data poisoning attacks",
"remediation": "Implement data integrity validation, anomaly detection in training pipelines, and input sanitization to protect against data poisoning.",
"priority": "HIGH",
},
"adversarial_protection": {
"label": "Protection against adversarial examples",
"remediation": "Test for adversarial vulnerabilities. Implement adversarial training, input validation, and detection mechanisms as appropriate.",
"priority": "HIGH",
},
"model_manipulation_protection": {
"label": "Protection against model manipulation (flipping)",
"remediation": "Implement model integrity verification, secure model storage, and access controls to prevent unauthorised model modifications.",
"priority": "MEDIUM",
},
"confidentiality_protection": {
"label": "Protection against confidentiality attacks",
"remediation": "Protect against model inversion, membership inference, and data extraction attacks. Implement differential privacy or other privacy-preserving techniques where appropriate.",
"priority": "MEDIUM",
},
"security_proportionate": {
"label": "Security measures proportionate to risks",
"remediation": "Conduct a cybersecurity risk assessment specific to the AI system. Ensure security measures are proportionate to the identified risks.",
"priority": "MEDIUM",
},
},
},
"quality_management": {
"article": "Art. 17",
"title": "Quality Management System",
"weight": 8,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"compliance_strategy": {
"label": "Regulatory compliance strategy defined",
"remediation": "Document a strategy for achieving and maintaining compliance with the AI Act, including roles, responsibilities, timelines, and resource allocation.",
"priority": "HIGH",
},
"design_verification": {
"label": "Design verification/validation techniques documented",
"remediation": "Document the verification and validation techniques used to confirm the system meets its design specifications and intended purpose.",
"priority": "MEDIUM",
},
"development_processes": {
"label": "Development, QC, and QA processes documented",
"remediation": "Document all development, quality control, and quality assurance processes including code review, testing protocols, and release management.",
"priority": "MEDIUM",
},
"test_procedures": {
"label": "Examination, test, and validation procedures documented",
"remediation": "Document pre-development, during-development, and post-development testing and validation procedures.",
"priority": "MEDIUM",
},
"standards_documented": {
"label": "Technical specifications and standards applied documented",
"remediation": "List all technical standards, harmonised standards, and common specifications applied in system development and operation.",
"priority": "LOW",
},
"data_management_systems": {
"label": "Data management systems and procedures in place",
"remediation": "Implement documented data management systems covering collection, analysis, labelling, storage, filtration, aggregation, retention, and disposal.",
"priority": "MEDIUM",
},
"risk_management_integrated": {
"label": "Risk management system (Art. 9) integrated into QMS",
"remediation": "Ensure the AI risk management system is formally integrated into the broader QMS, with clear interfaces and feedback loops.",
"priority": "HIGH",
},
"post_market_monitoring": {
"label": "Post-market monitoring (Art. 72) integrated into QMS",
"remediation": "Integrate post-market monitoring processes into the QMS with defined data flows, triggers for corrective action, and management review.",
"priority": "HIGH",
},
"incident_reporting_procedures": {
"label": "Incident and malfunction reporting procedures defined",
"remediation": "Define procedures for identifying, documenting, and reporting incidents and malfunctions, including the 15-day serious incident reporting deadline.",
"priority": "HIGH",
},
"authority_communication": {
"label": "Authority and notified body communication procedures defined",
"remediation": "Establish procedures for communicating with market surveillance authorities and notified bodies, including designated contacts and response timelines.",
"priority": "LOW",
},
"record_keeping_procedures": {
"label": "Record-keeping procedures defined",
"remediation": "Define procedures for creating, maintaining, and retaining QMS records including retention periods and access controls.",
"priority": "LOW",
},
"resource_management": {
"label": "Resource management including supply-chain measures defined",
"remediation": "Document resource management practices including personnel competency requirements, infrastructure needs, and supply-chain due diligence for AI components.",
"priority": "MEDIUM",
},
"accountability_framework": {
"label": "Accountability framework established",
"remediation": "Establish a clear accountability framework with defined roles, responsibilities, and reporting lines for AI Act compliance.",
"priority": "HIGH",
},
},
},
"conformity_assessment": {
"article": "Art. 43",
"title": "Conformity Assessment",
"weight": 8,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"path_determined": {
"label": "Assessment path determined (internal control vs third-party)",
"remediation": "Determine whether internal control (Annex VI) or third-party assessment (Annex VII) is required based on the system category and applicable harmonised standards.",
"priority": "HIGH",
},
"documentation_compiled": {
"label": "All required documentation compiled",
"remediation": "Compile the complete technical file including all documentation required for the chosen conformity assessment path.",
"priority": "HIGH",
},
"assessment_completed": {
"label": "Assessment procedure completed (Annex VI or VII)",
"remediation": "Complete the full conformity assessment procedure. For internal control: self-assess all requirements. For third-party: engage notified body and complete review.",
"priority": "HIGH",
},
"declaration_signed": {
"label": "EU Declaration of Conformity signed (Art. 47)",
"remediation": "Prepare and sign the EU Declaration of Conformity containing all required information per Art. 47. Keep available for 10 years after system is placed on market.",
"priority": "HIGH",
},
},
},
"ce_marking": {
"article": "Art. 48",
"title": "CE Marking",
"weight": 4,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"marking_affixed": {
"label": "CE marking affixed visibly, legibly, and indelibly",
"remediation": "Affix the CE marking to the AI system, its packaging, or accompanying documentation in a visible, legible, and indelible manner.",
"priority": "HIGH",
},
"notified_body_number": {
"label": "Notified body identification number included (if applicable)",
"remediation": "If a notified body was involved in conformity assessment, include its identification number alongside the CE marking.",
"priority": "MEDIUM",
},
"precedes_market_placement": {
"label": "CE marking precedes placing on market",
"remediation": "Ensure CE marking is applied before the system is placed on the market or put into service.",
"priority": "HIGH",
},
},
},
"eu_database": {
"article": "Art. 49",
"title": "EU Database Registration",
"weight": 5,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"provider_registered": {
"label": "Provider registered in EU database",
"remediation": "Register the provider entity in the EU AI database established under Art. 71.",
"priority": "HIGH",
},
"system_description_entered": {
"label": "System description and intended purpose entered",
"remediation": "Enter a description of the AI system and its intended purpose in the EU database.",
"priority": "HIGH",
},
"conformity_status_entered": {
"label": "Conformity assessment status entered",
"remediation": "Enter the conformity assessment status and any relevant certificate numbers.",
"priority": "MEDIUM",
},
"member_states_entered": {
"label": "Member states of availability entered",
"remediation": "Enter the list of EU member states where the system is or will be made available.",
"priority": "MEDIUM",
},
"restrictions_documented": {
"label": "Any usage restrictions documented",
"remediation": "Document any restrictions or conditions on the use of the system in the database entry.",
"priority": "LOW",
},
},
},
"post_market_monitoring": {
"article": "Art. 72",
"title": "Post-Market Monitoring",
"weight": 7,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"system_established": {
"label": "Post-market monitoring system established",
"remediation": "Establish a documented post-market monitoring system proportionate to the system's risks. Define responsibilities, data sources, and procedures.",
"priority": "HIGH",
},
"data_collection_defined": {
"label": "Data collection and analysis procedures defined",
"remediation": "Define procedures for actively collecting, documenting, and analysing relevant performance and safety data from deployers and other sources.",
"priority": "HIGH",
},
"risk_management_updated": {
"label": "Findings feed into risk management updates",
"remediation": "Establish a formal feedback loop from post-market monitoring findings to risk management system updates.",
"priority": "HIGH",
},
"plan_in_documentation": {
"label": "Post-market monitoring plan included in technical documentation",
"remediation": "Include the post-market monitoring plan as part of the technical documentation file.",
"priority": "MEDIUM",
},
"incident_reporting_defined": {
"label": "Serious incident reporting procedures defined (Art. 73, 15-day deadline)",
"remediation": "Define procedures for reporting serious incidents to market surveillance authorities within 15 days. Include escalation criteria, report templates, and responsible persons.",
"priority": "HIGH",
},
"interaction_analysis": {
"label": "Analysis of interaction with other AI systems included",
"remediation": "Include analysis of how the AI system interacts with other AI systems in the post-market monitoring plan.",
"priority": "MEDIUM",
},
},
},
}
DEPLOYER_OBLIGATION_SECTION = {
"deployer": {
"article": "Art. 26",
"title": "Deployer Obligations",
"weight": 10,
"deadline": "2 Dec 2027 (Annex III) / 2 Aug 2028 (Annex I products)",
"fields": {
"use_per_instructions": {
"label": "System used according to provider's instructions",
"remediation": "Review and follow the provider's instructions for use. Train all operators on proper system usage. Document any deviations.",
"priority": "HIGH",
},
"human_oversight_assigned": {
"label": "Human oversight assigned to competent, trained persons",
"remediation": "Assign human oversight responsibilities to persons with appropriate competence, training, and authority. Provide role-specific training.",
"priority": "HIGH",
},
"input_data_relevant": {
"label": "Input data is relevant and sufficiently representative",
"remediation": "Validate that the input data used in operation is relevant to the system's intended purpose and representative of the target context.",
"priority": "MEDIUM",
},
"monitoring_active": {
"label": "System operation monitored; provider informed of risks/incidents",
"remediation": "Implement operational monitoring based on the provider's instructions. Establish procedures for informing the provider of any risks, incidents, or malfunctions.",
"priority": "HIGH",
},
"logs_retained": {
"label": "Automatically generated logs retained (minimum 6 months)",
"remediation": "Retain all automatically generated system logs for the required period (minimum 6 months unless otherwise specified by law). Implement log storage with appropriate access controls.",
"priority": "MEDIUM",
},
"workers_informed": {
"label": "Workers and representatives informed before deployment",
"remediation": "Inform all workers and their representatives who will be subject to the AI system before it is put into service. Document the notification.",
"priority": "MEDIUM",
},
"dpia_completed": {
"label": "Data Protection Impact Assessment completed (when required)",
"remediation": "Assess whether a DPIA is required under GDPR Art. 35. If so, complete and document the DPIA before deploying the AI system.",
"priority": "HIGH",
},
"fria_completed": {
"label": "Fundamental Rights Impact Assessment completed (public bodies)",
"remediation": "If the deployer is a public body or provides public services, complete a Fundamental Rights Impact Assessment per Art. 27. Submit results to the market surveillance authority.",
"priority": "HIGH",
},
"incidents_reported": {
"label": "Serious incidents reported to provider and authorities",
"remediation": "Establish procedures for reporting serious incidents to the provider and market surveillance authority. Train staff on incident identification and reporting.",
"priority": "HIGH",
},
},
},
}
GPAI_OBLIGATION_SECTION = {
"gpai": {
"article": "Art. 53",
"title": "GPAI Model Obligations",
"weight": 10,
"deadline": "2 Aug 2025",
"fields": {
"technical_docs": {
"label": "Technical documentation of model and training process maintained",
"remediation": "Create and maintain technical documentation per Annex XI covering model architecture, training methodology, data sources, compute resources, and evaluation results.",
"priority": "HIGH",
},
"downstream_info": {
"label": "Information provided for downstream AI system providers",
"remediation": "Provide downstream providers with sufficient information about the model's capabilities, limitations, and proper integration to enable their own compliance.",
"priority": "HIGH",
},
"copyright_policy": {
"label": "Copyright compliance policy in place",
"remediation": "Implement a policy to comply with EU copyright law (Directive 2019/790), including mechanisms for rights holders to exercise text and data mining opt-out rights.",
"priority": "HIGH",
},
"training_summary_public": {
"label": "Training data summary publicly available",
"remediation": "Prepare and publish a sufficiently detailed summary of training data content per the AI Office template.",
"priority": "HIGH",
},
"eu_representative": {
"label": "EU authorised representative appointed (non-EU providers)",
"remediation": "Appoint an authorised representative established in the EU if the provider is based outside the EU.",
"priority": "HIGH",
},
},
},
}
GPAI_SYSTEMIC_SECTION = {
"gpai_systemic": {
"article": "Art. 55",
"title": "GPAI Systemic Risk Obligations",
"weight": 10,
"deadline": "2 Aug 2025",
"fields": {
"model_evaluation": {
"label": "Standardized model evaluation performed including adversarial testing",
"remediation": "Perform model evaluations using standardized protocols and benchmarks. Include adversarial testing scenarios proportionate to the risk level.",
"priority": "HIGH",
},
"red_teaming": {
"label": "Adversarial red-teaming conducted",
"remediation": "Conduct structured red-teaming exercises with qualified personnel to identify vulnerabilities, harmful outputs, and misuse potential.",
"priority": "HIGH",
},
"risk_assessment": {
"label": "Systemic risk assessment and mitigation performed",
"remediation": "Assess possible systemic risks at EU level (including their sources) and implement mitigation measures. Document the assessment methodology and findings.",
"priority": "HIGH",
},
"incident_tracking": {
"label": "Serious incidents tracked, documented, and reported to AI Office",
"remediation": "Implement incident tracking, documentation, and reporting procedures. Report serious incidents and corrective measures to the AI Office and national authorities.",
"priority": "HIGH",
},
"cybersecurity": {
"label": "Adequate cybersecurity for model and infrastructure ensured",
"remediation": "Implement cybersecurity protections for the model (against adversarial attacks, data poisoning) and its physical infrastructure.",
"priority": "HIGH",
},
"energy_reporting": {
"label": "Energy consumption reported",
"remediation": "Report the model's energy consumption for training and, where technically feasible, inference. Include overall energy efficiency metrics.",
"priority": "MEDIUM",
},
},
},
}
# ---------------------------------------------------------------------------
# Compliance checking logic
# ---------------------------------------------------------------------------
def check_section(section_def: Dict, status: Dict) -> Dict[str, Any]:
"""Check compliance for a single obligation section."""
total = len(section_def["fields"])
compliant = 0
gaps = []
met = []
for field_id, field_def in section_def["fields"].items():
value = status.get(field_id, False)
if value:
compliant += 1
met.append({
"item": field_def["label"],
"status": "COMPLIANT",
})
else:
gaps.append({
"item": field_def["label"],
"status": "GAP",
"priority": field_def["priority"],
"remediation": field_def["remediation"],
})
score = (compliant / total * 100) if total > 0 else 0
return {
"article": section_def["article"],
"title": section_def["title"],
"deadline": section_def["deadline"],
"weight": section_def["weight"],
"total_items": total,
"compliant_items": compliant,
"section_score": round(score, 1),
"gaps": gaps,
"met": met,
}
def calculate_overall_score(section_results: List[Dict]) -> float:
"""Calculate weighted overall compliance score (0-100)."""
total_weight = sum(s["weight"] for s in section_results)
if total_weight == 0:
return 0.0
weighted_sum = sum(s["section_score"] * s["weight"] for s in section_results)
return round(weighted_sum / total_weight, 1)
def determine_readiness(score: float) -> Dict[str, str]:
"""Determine compliance readiness level from score."""
if score >= 90:
return {
"level": "READY",
"description": "System meets compliance requirements. Minor items may remain. Proceed to conformity assessment.",
"color": "GREEN",
}
elif score >= 70:
return {
"level": "MOSTLY_READY",
"description": "System substantially compliant with some gaps. Address remaining items before conformity assessment.",
"color": "YELLOW",
}
elif score >= 50:
return {
"level": "PARTIALLY_READY",
"description": "Significant compliance gaps exist. Prioritize high-priority remediation items.",
"color": "ORANGE",
}
elif score >= 25:
return {
"level": "EARLY_STAGE",
"description": "Compliance program in early stages. Major effort required across multiple obligation areas.",
"color": "RED",
}
else:
return {
"level": "NOT_STARTED",
"description": "Compliance program not yet established. Begin with AI system inventory and risk classification.",
"color": "RED",
}
def run_compliance_check(config: Dict[str, Any]) -> Dict[str, Any]:
"""Run full compliance check against the provided configuration."""
result = {
"system_name": config.get("system_name", "Unknown"),
"risk_level": config.get("risk_level", "HIGH"),
"role": config.get("role", "provider"),
"check_date": datetime.now().strftime("%Y-%m-%d"),
"sections": [],
"overall_score": 0.0,
"readiness": {},
"priority_gaps": [],
"summary": {},
}
role = config.get("role", "provider").lower()
risk_level = config.get("risk_level", "HIGH").upper()
# Determine which sections to check
sections_to_check = {}
if role == "provider" and risk_level in ("HIGH", "UNACCEPTABLE"):
sections_to_check.update(PROVIDER_OBLIGATION_SECTIONS)
if role == "deployer":
sections_to_check.update(DEPLOYER_OBLIGATION_SECTION)
if config.get("is_gpai"):
sections_to_check.update(GPAI_OBLIGATION_SECTION)
if config.get("is_systemic_risk"):
sections_to_check.update(GPAI_SYSTEMIC_SECTION)
# If provider also has deployer obligations flagged
if role == "provider" and "deployer" in config:
sections_to_check.update(DEPLOYER_OBLIGATION_SECTION)
# Run checks
section_results = []
all_gaps = []
for section_id, section_def in sections_to_check.items():
status = config.get(section_id, {})
section_result = check_section(section_def, status)
section_result["section_id"] = section_id
section_results.append(section_result)
for gap in section_result["gaps"]:
gap["section"] = section_result["title"]
gap["article"] = section_result["article"]
gap["deadline"] = section_result["deadline"]
all_gaps.append(gap)
result["sections"] = section_results
# Calculate overall score
result["overall_score"] = calculate_overall_score(section_results)
result["readiness"] = determine_readiness(result["overall_score"])
# Sort gaps by priority
priority_order = {"HIGH": 0, "MEDIUM": 1, "LOW": 2}
all_gaps.sort(key=lambda g: priority_order.get(g["priority"], 3))
result["priority_gaps"] = all_gaps
# Summary
total_items = sum(s["total_items"] for s in section_results)
compliant_items = sum(s["compliant_items"] for s in section_results)
result["summary"] = {
"total_obligations_checked": total_items,
"obligations_met": compliant_items,
"obligations_gaps": total_items - compliant_items,
"high_priority_gaps": len([g for g in all_gaps if g["priority"] == "HIGH"]),
"medium_priority_gaps": len([g for g in all_gaps if g["priority"] == "MEDIUM"]),
"low_priority_gaps": len([g for g in all_gaps if g["priority"] == "LOW"]),
"sections_checked": len(section_results),
}
return result
# ---------------------------------------------------------------------------
# Output formatting
# ---------------------------------------------------------------------------
def format_text_report(result: Dict[str, Any]) -> str:
"""Format compliance check result as a human-readable report."""
lines = []
lines.append("=" * 72)
lines.append("EU AI ACT COMPLIANCE CHECK REPORT")
lines.append("Regulation (EU) 2024/1689")
lines.append("=" * 72)
lines.append("")
lines.append(f"System: {result['system_name']}")
lines.append(f"Risk Level: {result['risk_level']}")
lines.append(f"Role: {result['role'].title()}")
lines.append(f"Check Date: {result['check_date']}")
lines.append(f"Overall Score: {result['overall_score']}%")
lines.append(f"Readiness: {result['readiness']['level']} ({result['readiness']['color']})")
lines.append(f" {result['readiness']['description']}")
lines.append("")
# Summary
s = result["summary"]
lines.append("-" * 72)
lines.append("SUMMARY")
lines.append("-" * 72)
lines.append(f" Obligations checked: {s['total_obligations_checked']}")
lines.append(f" Obligations met: {s['obligations_met']}")
lines.append(f" Gaps found: {s['obligations_gaps']}")
lines.append(f" High priority: {s['high_priority_gaps']}")
lines.append(f" Medium priority: {s['medium_priority_gaps']}")
lines.append(f" Low priority: {s['low_priority_gaps']}")
lines.append("")
# Section scores
lines.append("-" * 72)
lines.append("SECTION SCORES")
lines.append("-" * 72)
for sec in result["sections"]:
bar_filled = int(sec["section_score"] / 5)
bar_empty = 20 - bar_filled
bar = "#" * bar_filled + "." * bar_empty
lines.append(f" [{bar}] {sec['section_score']:5.1f}% {sec['title']} ({sec['article']})")
lines.append("")
# Priority gaps
if result["priority_gaps"]:
lines.append("-" * 72)
lines.append("GAP ANALYSIS (sorted by priority)")
lines.append("-" * 72)
current_priority = None
for gap in result["priority_gaps"]:
if gap["priority"] != current_priority:
current_priority = gap["priority"]
lines.append(f"\n === {current_priority} PRIORITY ===")
lines.append(f"\n [{gap['article']}] {gap['section']}")
lines.append(f" Gap: {gap['item']}")
lines.append(f" Deadline: {gap['deadline']}")
lines.append(f" Remediation: {gap['remediation']}")
lines.append("")
# Section details
lines.append("-" * 72)
lines.append("DETAILED SECTION RESULTS")
lines.append("-" * 72)
for sec in result["sections"]:
lines.append(f"\n {sec['title']} ({sec['article']}) — {sec['section_score']}%")
lines.append(f" Deadline: {sec['deadline']}")
lines.append(f" {sec['compliant_items']}/{sec['total_items']} items compliant")
for item in sec["met"]:
lines.append(f" [PASS] {item['item']}")
for gap in sec["gaps"]:
lines.append(f" [FAIL] {gap['item']}")
lines.append("")
lines.append("=" * 72)
lines.append("End of Report")
lines.append("=" * 72)
return "\n".join(lines)
# ---------------------------------------------------------------------------
# CLI
# ---------------------------------------------------------------------------
def main():
parser = argparse.ArgumentParser(
description="EU AI Act Compliance Checker — Validate compliance against Regulation (EU) 2024/1689",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog="""
Examples:
%(prog)s --input compliance_status.json
%(prog)s --input compliance_status.json --json
%(prog)s --input compliance_status.json --role deployer
%(prog)s --input compliance_status.json --output report.json
""",
)
parser.add_argument("--input", "-i", required=True, help="Path to JSON file with compliance status")
parser.add_argument("--output", "-o", help="Path to write output report (default: stdout)")
parser.add_argument("--json", action="store_true", help="Output in JSON format")
parser.add_argument("--role", choices=["provider", "deployer"], help="Override role (default: from input)")
args = parser.parse_args()
try:
with open(args.input, "r", encoding="utf-8") as f:
config = json.load(f)
except FileNotFoundError:
print(f"Error: File not found — {args.input}", file=sys.stderr)
sys.exit(1)
except json.JSONDecodeError as e:
print(f"Error: Invalid JSON in {args.input} — {e}", file=sys.stderr)
sys.exit(1)
if args.role:
config["role"] = args.role
result = run_compliance_check(config)
if args.json:
output = json.dumps(result, indent=2, ensure_ascii=False)
else:
output = format_text_report(result)
if args.output:
with open(args.output, "w", encoding="utf-8") as f:
f.write(output)
print(f"Report written to {args.output}")
else:
print(output)
if __name__ == "__main__":
main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |