AI System Classification Standard

Document Type: Standard
Standard ID: STD-AI-001
Standard Title: AI System Classification Standard
Version: 1.0
Effective Date: 2025-08-01
Next Review Date: 2026-08-01
Review Frequency: Annually or upon regulatory change
Parent Policy: POL-AI-001 - Artificial Intelligence Policy
Owner: AI Act Program Manager
Approved By: AI Governance Committee Chair
Status: Draft
Classification: Internal Use Only

---

TABLE OF CONTENTS

1. Document History
2. Objective
3. Scope and Applicability
4. Control Standard
5. Supporting Procedures
6. Compliance
7. Roles and Responsibilities
8. Exceptions
9. Enforcement
10. Key Performance Indicators (KPIs)
11. Training Requirements
12. Definitions
13. Link with AI Act and ISO42001

---

DOCUMENT HISTORY

Version	Date	Author	Changes	Approval Date	Approved By
0.1	2025-06-15	Sarah Johnson, AI Act Program Manager	Initial draft	-	-
0.2	2025-07-01	Sarah Johnson, AI Act Program Manager	Incorporated stakeholder feedback	-	-
0.3	2025-07-20	Sarah Johnson, AI Act Program Manager	Added Annex III details	-	-
1.0	2025-08-01	Sarah Johnson, AI Act Program Manager	Final version approved	2025-07-25	Jane Doe, AI Governance Committee Chair

---

OBJECTIVE

This standard defines the requirements and process for classifying AI systems according to EU AI Act risk categories (prohibited, high-risk, limited-risk, minimal-risk) to ensure appropriate compliance obligations are applied to each AI system throughout its lifecycle.

Primary Goals:

• Prevent deployment of prohibited AI practices
• Accurately identify high-risk AI systems requiring full EU AI Act compliance
• Maintain a comprehensive classification register for all organizational AI systems
• Ensure timely reclassification when AI systems or regulations change

---

SCOPE AND APPLICABILITY

2.1 Mandatory Applicability

This standard is mandatory for:

• All AI systems developed, deployed, or distributed by the organization
• All AI systems at any lifecycle stage (design, development, testing, deployment, operation, decommissioning)
• All business units and third parties developing or using AI systems on behalf of the organization
• All AI system modifications that may affect classification

2.2 AI Systems Covered

This standard applies to all systems meeting the EU AI Act definition of "AI system" including:

• Machine learning systems (supervised, unsupervised, reinforcement learning)
• Logic- and knowledge-based systems (expert systems, rule-based systems)
• Statistical approaches (Bayesian methods, statistical inference)
• Search and optimization methods (genetic algorithms, particle swarm optimization)
• Any system meeting EU AI Act Article 3 definition of "AI system"

2.3 Geographical Scope

• All AI systems placed on the EU market
• All AI systems used within the EU regardless of provider location
• All AI systems with EU users or affecting EU persons

2.4 Out of Scope

• AI systems used exclusively for personal, non-professional activity
• AI systems used exclusively for military, defense, or national security purposes
• AI systems used exclusively for research and development (pre-market)

---

CONTROL STANDARD

Control CLS-001: Prohibited Practice Screening

Control ID: CLS-001
Control Name: Prohibited Practice Assessment and Prevention
Control Type: Preventive
Control Frequency: Before development starts, before deployment, annually
Risk Level: Critical

Control Objective

Ensure no AI systems constitute prohibited practices under EU AI Act Article 5 by systematically screening all AI systems against the 8 prohibited practices before development or deployment, thereby preventing legal violations, regulatory penalties, and reputational damage.

Control Requirements

CR-001.1: Prohibited Practice Screening Process

All AI systems must be screened against Article 5 prohibited practices using the approved assessment form before any development or deployment activities commence.

Mandatory Actions:

• Complete Prohibited Practice Assessment Form (FORM-AI-CLS-001) for each AI system
• Screen against all 8 prohibited practices defined in Article 5(1)(a-h)
• Document assessment results with clear justification
• Halt all development/deployment if any prohibited practice is identified
• Obtain legal review for borderline or unclear cases

Article 5 Prohibited Practices:

Practice	Article	Description	Assessment Criteria
1. Subliminal Techniques	5(1)(a)	Deploying subliminal techniques beyond person's consciousness OR purposefully manipulative or deceptive techniques to materially distort behavior causing significant harm	Does AI use subliminal manipulation? Does it bypass conscious awareness? Does it use manipulative or deceptive techniques? Could it cause harm?
2. Exploitation of Vulnerabilities	5(1)(b)	Exploiting vulnerabilities due to age, disability, social/economic situation to materially distort behavior causing harm	Does AI target vulnerable groups? Does it exploit vulnerabilities? Could it cause harm?
3. Social Scoring	5(1)(c)	Evaluation/classification of persons over a certain period of time based on their social behaviour or known, inferred or predicted personal or personality characteristics, leading to either (i) detrimental treatment in contexts unrelated to the original data collection context, or (ii) treatment that is unjustified or disproportionate	Is it social scoring? Is it over a certain period of time? Based on social behavior or personal/personality characteristics? Does the score lead to detrimental treatment in an unrelated context OR unjustified/disproportionate treatment?
4. Crime Risk Profiling	5(1)(d)	Assessing risk of natural person committing criminal offense based solely on profiling or personality traits (not objective criminal activity data)	Does it assess crime risk? Based only on profiling? Not based on objective criminal data?
5. Untargeted Facial Scraping	5(1)(e)	Creating or expanding facial recognition databases through untargeted scraping from internet or CCTV	Does it scrape facial images? Is scraping untargeted? From internet/CCTV?
6. Emotion Recognition in Workplace/Education	5(1)(f)	Inferring emotions in workplace or educational institutions (except medical or safety reasons)	Does it infer emotions? In workplace or education setting? Not for medical/safety?
7. Biometric Categorization (Sensitive Attributes)	5(1)(g)	Categorizing persons based on biometric data to infer sensitive attributes (race, political opinions, religion, etc.)	Does it categorize using biometrics? Does it infer sensitive attributes?
8. Real-Time Biometric ID in Public	5(1)(h)	Real-time remote biometric identification in publicly accessible spaces for law enforcement (limited exceptions)	Real-time biometric ID? In public spaces? For law enforcement?

Evidence Required:

• Completed FORM-AI-CLS-001 (Prohibited Practice Assessment Form)
• Legal review records for borderline cases
• Assessment approval by AI Act Program Manager
• Halt decision documentation if prohibited practice identified

Audit Verification:

• Review 100% of AI systems for prohibited practice assessment completion
• Verify assessment quality through sampling (minimum 20%)
• Confirm legal review for all borderline cases
• Verify no prohibited practices are in development or operation

---

Control CLS-002: High-Risk Classification (Annex I - Product Safety)

Control ID: CLS-002
Control Name: Annex I Product Safety AI Assessment
Control Type: Preventive
Control Frequency: Initial classification, after substantial modifications
Risk Level: High

Control Objective

Accurately identify AI systems that are safety components of products covered by Union harmonization legislation (Annex I) to ensure they undergo appropriate conformity assessment and meet product safety requirements.

Control Requirements

CR-002.1: Annex I Product Assessment

Assess whether the AI system is a safety component of a product listed in Annex I and subject to third-party conformity assessment under Union harmonization legislation.

Mandatory Actions:

• Review AI system against all 12 Annex I product categories
• Determine if AI system is a "safety component" of the product
• Assess if product requires third-party conformity assessment
• Complete Annex I Assessment Form (FORM-AI-CLS-002)
• Classify as HIGH-RISK (Annex I) if all three criteria are met

Annex I Product Categories:

Category	Legislation	Assessment Question
Machinery	Machinery Regulation	Is AI a safety component of machinery requiring conformity assessment?
Toys	Toys Directive	Is AI a safety component of toys requiring conformity assessment?
Recreational craft and personal watercraft	Recreational Craft Directive	Is AI a safety component requiring conformity assessment?
Lifts	Lifts Directive	Is AI a safety component of lifts requiring conformity assessment?
Equipment in explosive atmospheres	ATEX Directive	Is AI a safety component of ATEX equipment requiring conformity assessment?
Radio equipment	Radio Equipment Directive	Is AI a safety component of radio equipment requiring conformity assessment?
Pressure equipment	Pressure Equipment Directive	Is AI a safety component of pressure equipment requiring conformity assessment?
Cableway installations	Cableway Directive	Is AI a safety component of cableways requiring conformity assessment?
Personal protective equipment	PPE Regulation	Is AI a safety component of PPE requiring conformity assessment?
Appliances burning gaseous fuels	Gas Appliances Regulation	Is AI a safety component of gas appliances requiring conformity assessment?
Medical devices	Medical Device Regulation	Is AI a safety component of medical devices requiring conformity assessment?
In vitro diagnostic medical devices	IVD Regulation	Is AI a safety component of IVD devices requiring conformity assessment?

Three-Part Test for Annex I Classification:

1. ✓ Is the AI system part of a product listed in Annex I?
2. ✓ Is the AI system a safety component of that product?
3. ✓ Does the product require third-party conformity assessment?

If YES to all three → HIGH-RISK (Annex I)

Evidence Required:

• Completed FORM-AI-CLS-002 (Annex I Assessment Form)
• Product documentation and specifications
• Safety component determination with justification
• Conformity assessment requirement verification

Audit Verification:

• Verify assessment completion for all AI systems that are product components
• Validate safety component determination through technical review
• Confirm conformity assessment requirements correctly identified
• Cross-check with product safety documentation

---

Control CLS-003: High-Risk Classification (Annex III - Standalone Use Cases)

Control ID: CLS-003
Control Name: Annex III Standalone High-Risk AI Assessment
Control Type: Preventive
Control Frequency: Initial classification, after substantial modifications
Risk Level: High

Control Objective

Accurately identify AI systems that fall under Annex III high-risk use cases to ensure they comply with all EU AI Act requirements for high-risk systems, protecting fundamental rights and safety of persons.

Control Requirements

CR-003.1: Annex III Use Case Assessment

Systematically assess whether the AI system falls under any of the 8 Annex III high-risk categories by evaluating the AI system's intended purpose, deployment context, and impact on persons.

Mandatory Actions:

• Review AI system against all 8 Annex III categories
• Assess each specific use case within applicable categories
• Complete Annex III Assessment Form (FORM-AI-CLS-003) with detailed justification
• Obtain legal review for borderline cases
• Classify as HIGH-RISK (Annex III) if any category applies

Annex III High-Risk Categories:

Category 1: Biometrics (Annex III #1)

Use Case	Description	Classification Trigger
Remote Biometric Identification	Real-time or post identification of persons	ANY remote biometric identification system
Biometric Categorization	Categorizing persons based on biometric data to infer sensitive attributes	ANY biometric categorization for sensitive attributes
Emotion Recognition	Inferring emotions from biometric data	ANY emotion recognition system

Category 2: Critical Infrastructure (Annex III #2)

Use Case	Description	Classification Trigger
Critical Digital Infrastructure	AI as safety component in critical digital infrastructure management/operation	AI controls/manages critical digital infrastructure AND is safety component
Road Traffic Management	AI as safety component in road traffic management/operation	AI controls/manages road traffic AND is safety component
Utilities Supply	AI as safety component in water, gas, heating, electricity supply	AI controls/manages utilities AND is safety component

Category 3: Education and Vocational Training (Annex III #3)

Use Case	Description	Classification Trigger
Educational Access	Determining access or admission to educational institutions	AI makes/influences admission decisions
Educational Assignment	Assigning persons to educational institutions	AI assigns students to schools/programs
Learning Outcome Evaluation	Evaluating learning outcomes or educational performance	AI evaluates student performance/outcomes
Exam Proctoring	Monitoring and detecting prohibited behavior during tests	AI monitors/detects cheating or violations
Education Level Assessment	Assessing appropriate level of education for person	AI determines educational level/placement

Category 4: Employment, Workers Management, and Self-Employment (Annex III #4)

Use Case	Description	Classification Trigger
Recruitment	Targeted job advertisements, screening, evaluation of candidates	AI screens CVs, evaluates candidates, targets job ads
Employment Decisions	Making decisions on hiring, promotion, termination, task allocation	AI makes/influences hire/fire/promote decisions
Performance Monitoring	Monitoring and evaluating performance and behavior of workers	AI monitors worker performance or behavior

Category 5: Access to Essential Private/Public Services and Benefits (Annex III #5)

Use Case	Description	Classification Trigger
Public Assistance Eligibility	Evaluating eligibility for public assistance benefits and services	AI determines eligibility for public benefits
Creditworthiness Assessment	Evaluating creditworthiness (except fraud detection)	AI assesses credit risk or loan eligibility
Emergency Service Dispatching	Dispatching or establishing priority for emergency first response services	AI dispatches ambulance/fire/police services
Insurance Risk Assessment	Risk assessment and pricing for life and health insurance	AI assesses risk or sets premiums

Category 6: Law Enforcement (Annex III #6)

Use Case	Description	Classification Trigger
Crime Victim Risk Assessment	Assessing risk of person becoming victim of criminal offense	AI predicts victimization risk
Polygraph Tools	Polygraphs and similar tools to detect emotional state	AI detects deception or emotional state
Evidence Reliability	Evaluating reliability of evidence in criminal proceedings	AI assesses evidence credibility
Crime Risk Prediction	Assessing risk of person committing or reoffending based on profiling	AI predicts criminal behavior via profiling
Criminal Profiling	Profiling in detection, investigation, or prosecution of criminal offenses	AI profiles suspects or criminal patterns

Category 7: Migration, Asylum, and Border Control (Annex III #7)

Use Case	Description	Classification Trigger
Border Polygraph Tools	Polygraphs and similar tools for border/immigration control	AI detects deception at borders
Border Risk Assessment	Assessing risk posed by persons entering territory	AI assesses security risk of entrants
Immigration Application Examination	Examining applications for asylum, visa, residence permits	AI evaluates/decides on immigration applications
Border Person Identification	Detecting, recognizing, or identifying persons at borders (except verification)	AI identifies persons crossing borders

Category 8: Administration of Justice and Democratic Processes (Annex III #8)

Use Case	Article	Description	Classification Trigger
Judicial and ADR Assistance	8(a)	Assisting a judicial authority in researching and interpreting facts and the law AND in applying the law to a concrete set of facts, or used in a similar way in alternative dispute resolution	AI assists judges/arbitrators in legal research, law interpretation, law application to facts, or alternative dispute resolution
Election/Voting Influence	8(b)	Influencing the outcome of an election or referendum, or the voting behaviour of natural persons, including the use of AI systems to influence political campaigns	AI influences election outcomes, referendum results, or voting behaviour

Evidence Required:

• Completed FORM-AI-CLS-003 (Annex III Assessment Form)
• Use case description and intended purpose documentation
• Deployment context analysis
• Legal review records for borderline cases
• Classification decision with justification

Audit Verification:

• Review 100% of AI systems for Annex III assessment completion
• Verify assessment accuracy through sampling (minimum 20%)
• Validate use case categorization through expert review
• Confirm legal review for all borderline determinations

---

Control CLS-004: Classification Decision and Documentation

Control ID: CLS-004
Control Name: Final Classification Determination and Documentation
Control Type: Preventive
Control Frequency: Per AI system, after substantial modifications
Risk Level: High

Control Objective

Make definitive classification decisions based on comprehensive assessments and maintain complete documentation to ensure traceability, support audits, and enable appropriate compliance obligations for each AI system.

Control Requirements

CR-004.1: Classification Decision Tree Execution

Apply the standardized classification decision tree to make final classification determination based on results from prohibited practice, Annex I, and Annex III assessments.

Classification Decision Tree:

STEP 1: Prohibited Practice Check (Article 5)
├─ Is AI system a prohibited practice?
   ├─ YES → Classification: PROHIBITED
   │         Action: Do NOT develop or deploy
   │         Escalation: IMMEDIATE to AI Governance Committee + Legal
   └─ NO → Continue to Step 2

STEP 2: Annex I Assessment (Product Safety)
├─ Is AI a safety component of Annex I product requiring conformity assessment?
   ├─ YES → Classification: HIGH-RISK (Annex I)
   │         Action: Apply ALL high-risk requirements + Annex VII conformity assessment
   │         Trigger: STD-AI-002 through STD-AI-014
   └─ NO → Continue to Step 3

STEP 3: Annex III Assessment (Standalone Use Cases)
├─ Does AI fall under any Annex III category?
   ├─ YES → Classification: HIGH-RISK (Annex III)
   │         Action: Apply ALL high-risk requirements + Annex VI conformity assessment
   │         Trigger: STD-AI-002 through STD-AI-014
   └─ NO → Continue to Step 4

STEP 4: Transparency Obligations (Article 50)
├─ Does AI have transparency obligations?
│  - Chatbot or conversational AI?
│  - Generates synthetic content (deepfakes)?
│  - Emotion recognition (not workplace/education)?
│  - Biometric categorization (not sensitive attributes)?
   ├─ YES → Classification: LIMITED-RISK
   │         Action: Apply transparency disclosure requirements
   │         Trigger: STD-AI-006 (Transparency Standard)
   └─ NO → Continue to Step 5

STEP 5: Default Classification
└─ Classification: MINIMAL-RISK
   Action: No specific EU AI Act obligations (voluntary best practices recommended)

Mandatory Actions:

• Execute decision tree systematically for each AI system
• Document decision at each step with justification
• Complete Classification Decision Record (REC-AI-CLS-001)
• Obtain AI Governance Committee approval for HIGH-RISK classifications
• Trigger applicable compliance obligations based on classification

Evidence Required:

• Classification Decision Record (REC-AI-CLS-001)
• All assessment forms (FORM-AI-CLS-001, -002, -003)
• Decision tree execution documentation
• AI Governance Committee approval for high-risk systems
• Classification certificate

Audit Verification:

• Verify decision tree properly executed for 100% of AI systems
• Validate classification decisions through sampling
• Confirm AI Governance Committee approvals obtained
• Verify compliance obligations correctly triggered

---

Control CLS-005: Classification Register Maintenance

Control ID: CLS-005
Control Name: AI System Classification Register Management
Control Type: Detective
Control Frequency: Continuous updates, monthly reviews
Risk Level: Medium

Control Objective

Maintain a comprehensive, current, and accurate register of all AI system classifications to enable effective oversight, compliance monitoring, audit support, and regulatory reporting.

Control Requirements

CR-005.1: Classification Register Management

Maintain a centralized AI System Classification Register containing all classification decisions, statuses, and key metadata for every AI system in the organization.

Mandatory Register Fields:

Field	Description	Mandatory	Format/Values
AI_System_ID	Unique identifier	YES	AI-SYS-XXXX
AI_System_Name	Official name	YES	Text
Classification	Risk classification	YES	PROHIBITED / HIGH-RISK / LIMITED-RISK / MINIMAL-RISK
Classification_Basis	Legal basis for classification	YES	Article 5 / Annex I / Annex III / Article 50 / Default
Annex_III_Category	Specific Annex III category (if applicable)	If HIGH-RISK	#1-#8
Classification_Date	Date of classification decision	YES	YYYY-MM-DD
Classifier_Name	Person who performed classification	YES	Text
Approver_Name	Person who approved classification	YES	Text
Approval_Date	Date of approval	YES	YYYY-MM-DD
Next_Review_Date	Scheduled review date	YES	YYYY-MM-DD
Status	Current status	YES	Active / Under Review / Reclassified / Decommissioned
Lifecycle_Stage	Current development stage	YES	Design / Development / Testing / Deployed / Operational / Decommissioned
Business_Owner	Business unit owner	YES	Text
Compliance_Status	Compliance with classification requirements	YES	Compliant / Gaps Identified / Remediation In Progress

Mandatory Actions:

• Create register entry within 5 business days of classification decision
• Update register within 2 business days of any status change
• Conduct monthly register review for completeness and accuracy
• Generate quarterly register reports for AI Governance Committee
• Provide register access to internal audit and competent authorities
• Archive decommissioned AI system records (retain for 10 years)

Evidence Required:

• AI System Classification Register (REC-AI-CLS-002)
• Monthly register review records
• Quarterly reports to AI Governance Committee
• Register update audit trail
• Access logs for register

Audit Verification:

• Verify register completeness (100% of AI systems recorded)
• Validate register accuracy through sampling
• Confirm timely updates (within SLA)
• Review monthly review completion
• Verify 10-year retention for decommissioned systems

---

Control CLS-006: Periodic Classification Review

Control ID: CLS-006
Control Name: Annual Classification Review and Validation
Control Type: Detective
Control Frequency: Annually (minimum)
Risk Level: Medium

Control Objective

Ensure all AI system classifications remain accurate and current through systematic annual reviews, identifying any changes in AI systems, intended purposes, deployment contexts, or regulatory requirements that necessitate reclassification.

Control Requirements

CR-006.1: Annual Classification Review Process

Conduct comprehensive annual reviews of all AI system classifications to validate accuracy and identify reclassification needs.

Mandatory Actions:

• Schedule annual review for each AI system (based on last classification date)
• Re-assess against current Article 5, Annex I, and Annex III criteria
• Review for changes in: intended purpose, deployment context, functionality, user population
• Review for regulatory updates (EU AI Act guidance, implementing acts)
• Document review results in Annual Classification Review Record
• Update classification if changes identified
• Obtain approval for any reclassifications

Annual Review Checklist:

• Has the AI system's intended purpose changed?
• Has the deployment context changed?
• Has the user population changed?
• Have system capabilities or functionality changed?
• Have there been EU AI Act regulatory updates affecting classification?
• Has the AI system undergone substantial modifications?
• Is the prohibited practice assessment still valid?
• Is the Annex I/III assessment still valid?
• Does the classification still reflect the current risk level?

Evidence Required:

• Annual review schedule
• Annual Classification Review Records (per AI system)
• Updated classifications (if changed)
• AI Governance Committee approval for reclassifications
• Annual review completion report

Audit Verification:

• Verify 100% of AI systems reviewed annually
• Validate review quality through sampling
• Confirm reclassifications properly approved
• Verify timely completion of annual reviews

---

Control CLS-007: Event-Triggered Reclassification

Control ID: CLS-007
Control Name: Event-Triggered Classification Reassessment
Control Type: Detective
Control Frequency: As triggered by events
Risk Level: High

Control Objective

Ensure AI system classifications are reassessed immediately when triggering events occur that may affect the classification, preventing operation of AI systems under incorrect classifications and associated compliance gaps.

Control Requirements

CR-007.1: Reclassification Trigger Monitoring

Monitor for and respond to events that trigger mandatory reclassification assessment within defined timeframes.

Reclassification Triggers:

Trigger Event	Description	Action Required	Timeframe
Substantial Modification	Significant change to AI system affecting compliance or intended purpose	Immediate reclassification assessment	Within 10 business days
Intended Purpose Change	Change in how or why AI system is used	Immediate reclassification assessment	Within 10 business days
Deployment Context Change	Change in where, when, or on whom AI system is used	Reclassification assessment	Within 15 business days
Regulatory Update	EU AI Act amendments, implementing acts, or guidance affecting classification	Review impact, assess if reclassification needed	Within 30 days of publication
Regulatory Interpretation Change	New interpretation from Commission or competent authorities	Review impact, assess if reclassification needed	Within 30 days of notification
Incident or Near-Miss	Incident suggesting classification may be incorrect	Immediate reclassification assessment	Within 5 business days
Audit Finding	Internal/external audit identifies classification issue	Reclassification assessment	Within 10 business days
User Population Expansion	AI system used on new or broader user populations	Reclassification assessment	Within 15 business days

Mandatory Actions:

• Monitor systems and processes for reclassification triggers
• Log trigger event within 1 business day of identification
• Initiate reclassification assessment within required timeframe
• Conduct full classification assessment (all three: Article 5, Annex I, Annex III)
• Document reclassification decision
• Update Classification Register
• Communicate changes to stakeholders
• Halt deployment if reclassification increases risk level until compliance gaps addressed

Evidence Required:

• Trigger event log
• Trigger event documentation
• Reclassification assessment records
• Updated Classification Decision Record
• Updated Classification Register entry
• Stakeholder communication records
• Deployment hold records (if applicable)

Audit Verification:

• Verify all trigger events logged and assessed
• Validate reclassification completed within timeframes
• Confirm appropriate escalation and approval
• Verify stakeholder communications completed

---

Control CLS-008: Classification Change Management

Control ID: CLS-008
Control Name: Classification Change Control and Impact Management
Control Type: Corrective
Control Frequency: Per reclassification
Risk Level: High

Control Objective

Manage classification changes through formal change control to ensure compliance obligations are appropriately updated, gaps are identified and addressed, and all stakeholders are informed of classification changes and their implications.

Control Requirements

CR-008.1: Classification Change Control Process

Execute formal change control process for all classification changes, ensuring impact assessment, gap analysis, and remediation planning.

Mandatory Change Control Steps:

Step 1: Change Request (Day 0)

• Document reason for reclassification
• Identify current vs. new classification
• Submit Change Request (CR-AI-CLS-XXX)

Step 2: Impact Assessment (Days 1-3)

• Assess impact on compliance obligations
• Identify new requirements triggered by reclassification
• Identify requirements no longer applicable
• Estimate effort and resources required
• Document in Classification Change Impact Assessment

Step 3: Gap Analysis (Days 4-5)

• Compare current state to new classification requirements
• Identify compliance gaps
• Prioritize gaps by risk and regulatory deadline
• Document in Gap Analysis Report

Step 4: Remediation Planning (Days 6-10)

• Develop remediation plan for all gaps
• Assign owners and target dates
• Obtain resource commitments
• Define interim risk mitigation measures (if deployment continuing)

Step 5: Approval (Days 11-15)

• Present to AI Governance Committee
• Obtain approval for reclassification
• Obtain approval for remediation plan
• Obtain approval for interim risk mitigation (if needed)

Step 6: Implementation (Days 16+)

• Update Classification Register
• Implement remediation plan
• Update all affected documentation (technical docs, risk assessments, etc.)
• Communicate to all affected stakeholders
• Monitor remediation progress

Step 7: Verification (After remediation)

• Verify all gaps closed
• Confirm compliance with new classification requirements
• Obtain final sign-off

Classification Change Impact Matrix:

From → To	Impact Level	Key Actions Required
MINIMAL → LIMITED	Low	Implement transparency disclosures (STD-AI-006)
MINIMAL → HIGH-RISK	Critical	Full compliance program: STD-AI-002 through STD-AI-014
LIMITED → HIGH-RISK	High	Add all high-risk requirements beyond transparency
HIGH-RISK → LIMITED	Medium	Maintain documentation; reduce ongoing compliance burden
HIGH-RISK → MINIMAL	Medium	Maintain documentation; minimal ongoing obligations
Any → PROHIBITED	Critical	IMMEDIATE HALT; Do not deploy; Decommission if deployed

Evidence Required:

• Classification Change Request (CR-AI-CLS-XXX)
• Classification Change Impact Assessment
• Gap Analysis Report
• Remediation Plan
• AI Governance Committee approval records
• Updated Classification Register
• Stakeholder communication records
• Remediation completion verification

Audit Verification:

• Verify change control process followed for all reclassifications
• Validate impact assessment and gap analysis completed
• Confirm AI Governance Committee approval obtained
• Verify remediation completed before deployment (if upgrading to high-risk)
• Confirm stakeholder communications completed

---

Control CLS-009: Classification Documentation Package

Control ID: CLS-009
Control Name: Classification Documentation and Record Retention
Control Type: Detective
Control Frequency: Per classification, 10-year retention
Risk Level: Medium

Control Objective

Maintain complete, organized, and accessible documentation packages for all classification decisions to support audits, regulatory inspections, and demonstrate due diligence in classification processes.

Control Requirements

CR-009.1: Classification Documentation Package Management

Compile and retain complete documentation packages for each AI system classification decision for minimum 10 years.

Mandatory Documentation Package Contents:

Document	Document ID	Mandatory	Retention Period
Classification Request Form	FORM-AI-CLS-REQUEST-XXX	YES	10 years
Prohibited Practice Assessment	FORM-AI-CLS-001	YES	10 years
Annex I Assessment	FORM-AI-CLS-002	If product AI	10 years
Annex III Assessment	FORM-AI-CLS-003	YES	10 years
Legal Review (if conducted)	REVIEW-AI-CLS-LEGAL-XXX	If borderline case	10 years
Classification Decision Record	REC-AI-CLS-001	YES	10 years
AI Governance Committee Approval	APPROVAL-AI-CLS-XXX	If HIGH-RISK	10 years
Classification Certificate	CERT-AI-CLS-XXX	YES	10 years
Annual Review Records	REVIEW-AI-CLS-ANNUAL-XXX	YES (ongoing)	10 years after decommission
Reclassification Records	All reclassification docs	If reclassified	10 years
Change Impact Assessments	IMPACT-AI-CLS-XXX	If reclassified	10 years

Mandatory Actions:

• Create documentation package within 5 business days of classification decision
• Store in approved document management system with access controls
• Apply 10-year retention policy automatically
• Enable search and retrieval within 2 business days of request
• Provide to competent authorities within 5 business days of request
• Backup documentation packages in disaster recovery system
• Maintain audit trail of all access to documentation packages

Document Management Requirements:

• Secure storage with encryption at rest
• Role-based access control
• Version control for all documents
• Audit trail of all access and modifications
• Retention policy enforcement
• Disaster recovery and backup

Evidence Required:

• Complete documentation packages (one per AI system)
• Document management system records
• Retention policy configuration
• Access logs
• Backup verification records
• Authority request response records

Audit Verification:

• Verify 100% of AI systems have complete documentation packages
• Validate package completeness through sampling
• Confirm 10-year retention policy applied
• Verify access controls properly configured
• Test retrieval capability (sample requests)

---

SUPPORTING PROCEDURES

This standard is implemented through the following detailed procedures:

Procedure PROC-AI-CLS-001: AI System Classification Procedure

Purpose: Define step-by-step process for classifying new AI systems
Owner: AI Act Program Manager
Implements: Controls CLS-001, CLS-002, CLS-003, CLS-004, CLS-005, CLS-009

Procedure Steps:

1. Receive classification request from AI System Owner
2. Gather AI system information (intended purpose, functionality, users, deployment)
3. Conduct prohibited practice assessment (Control CLS-001)
4. Conduct Annex I assessment if AI is part of regulated product (Control CLS-002)
5. Conduct Annex III assessment (Control CLS-003)
6. Apply classification decision tree (Control CLS-004)
7. Obtain legal review for borderline cases
8. Document classification decision
9. Obtain AI Governance Committee approval (if HIGH-RISK)
10. Issue classification certificate
11. Update Classification Register (Control CLS-005)
12. Create documentation package (Control CLS-009)
13. Communicate classification to stakeholders
14. Trigger applicable compliance standards based on classification

Outputs:

• Classification certificate
• Updated Classification Register
• Documentation package
• Stakeholder communications

---

Procedure PROC-AI-CLS-002: Classification Review and Update Procedure

Purpose: Define process for periodic reviews and event-triggered reclassifications
Owner: AI Act Program Manager
Implements: Controls CLS-006, CLS-007, CLS-008

Procedure Steps:

1. For Annual Reviews (Control CLS-006):

   • Generate list of AI systems due for annual review
   • Schedule review meetings with AI System Owners
   • Conduct review using Annual Review Checklist
   • Re-assess against current Article 5, Annex I, Annex III criteria
   • Document review results
   • Initiate reclassification if needed

2. For Event-Triggered Reviews (Control CLS-007):

   • Monitor for reclassification trigger events
   • Log trigger event immediately
   • Assess if reclassification needed
   • Initiate reclassification within required timeframe
   • Document trigger and assessment

3. For Reclassifications (Control CLS-008):

   • Submit classification change request
   • Conduct impact assessment
   • Perform gap analysis
   • Develop remediation plan
   • Obtain AI Governance Committee approval
   • Update Classification Register
   • Implement remediation
   • Communicate changes to stakeholders
   • Verify remediation completion

Outputs:

• Annual review records
• Trigger event logs
• Reclassification assessments
• Updated classifications
• Gap analyses and remediation plans

---

Procedure PROC-AI-CLS-003: Classification Documentation and Record Management Procedure

Purpose: Define process for maintaining classification documentation packages
Owner: Document Control / AI Act Program Manager
Implements: Control CLS-009

Procedure Steps:

1. Create documentation package folder in document management system
2. Upload all required classification documents
3. Verify package completeness against checklist
4. Apply metadata (AI System ID, classification, dates)
5. Set retention policy (10 years)
6. Configure access permissions
7. Enable audit logging
8. Schedule backup
9. Conduct periodic documentation package audits (quarterly)
10. Respond to documentation requests from authorities

Outputs:

• Complete documentation packages
• Documentation audit reports
• Authority response records

---

COMPLIANCE

5.1 Compliance Monitoring

Monitoring Approach: Continuous automated monitoring supplemented by monthly manual reviews and quarterly comprehensive audits.

Compliance Metrics:

Metric	Target	Measurement Method	Frequency	Owner
Classification Completeness	100%	% of AI systems with current classification	Monthly	AI Act Program Manager
Classification Timeliness	< 10 business days	Average days from request to classification	Monthly	AI Act Program Manager
Annual Review Completion	100%	% of AI systems reviewed annually on schedule	Quarterly	AI Act Program Manager
Reclassification Responsiveness	100%	% of triggered reclassifications completed within SLA	Monthly	AI Act Program Manager
High-Risk Compliance	100%	% of high-risk AI with full compliance to all standards	Quarterly	Compliance Officer
Prohibited Practice Prevention	100% (Zero tolerance)	Zero prohibited practices in development or operation	Continuous	AI Governance Committee
Documentation Package Completeness	100%	% of AI systems with complete documentation packages	Monthly	Document Control
Authority Response Time	< 5 days	Average days to respond to authority documentation requests	Per request	Legal

Monitoring Tools:

• Classification Register Dashboard (real-time)
• Automated alerts for classification due dates
• Monthly compliance reports
• Quarterly AI Governance Committee reviews

---

5.2 Internal Audit Requirements

Audit Frequency: Annually (minimum)

Audit Scope:

• Classification Register completeness and accuracy
• Classification documentation quality and completeness
• Classification decision accuracy (sample-based testing)
• Annual review completion and quality
• Reclassification trigger response
• Compliance with classification procedures
• Controls effectiveness (CLS-001 through CLS-009)

Audit Activities:

• Review 100% of Classification Register for completeness
• Sample 20% of classifications for detailed accuracy review
• Test controls through transaction sampling
• Interview key personnel (classifiers, AI System Owners)
• Review AI Governance Committee meeting minutes
• Test documentation retrieval capability
• Verify retention policy enforcement

Audit Outputs:

• Annual Classification Audit Report
• Findings and recommendations
• Corrective action plans for deficiencies

---

5.3 External Audit / Regulatory Inspection

Preparation:

• Maintain audit-ready documentation packages at all times
• Designate Legal and AI Act Program Manager as regulatory liaisons
• Prepare standard response procedures for authority requests

Provide to Auditors/Regulators:

• AI System Classification Register (REC-AI-CLS-002)
• Sample classification documentation packages
• Classification procedures (PROC-AI-CLS-001, -002, -003)
• Annual review completion reports
• Internal audit reports
• AI Governance Committee meeting minutes (relevant to classification)
• Evidence of controls execution

Authority Request Response:

• Acknowledge request within 1 business day
• Provide requested documentation within 5 business days
• Coordinate through Legal and AI Act Program Manager
• Document all interactions with authorities

---

ROLES AND RESPONSIBILITIES

6.1 RACI Matrix

Activity	AI Act Program Manager	AI System Owner	Product Director	Legal & Compliance	AI Governance Committee	DPO	CTO
Classification Execution	R	A	C	C	I	C	I
Prohibited Practice Assessment	R	A	I	C	I	C	I
Annex I Assessment	C	R	A	C	I	I	A
Annex III Assessment	R	A	C	C	I	I	I
Classification Decision	R	A	I	C	I	I	I
High-Risk Approval	R	I	I	C	A	I	I
Classification Register Management	R/A	I	I	I	I	I	I
Annual Reviews	R	A	I	I	I	I	I
Reclassification Assessment	R	A	C	C	I	I	I
Legal Review (Borderline Cases)	C	I	I	R/A	I	C	I
Documentation Package Management	R	C	I	I	I	I	I
Compliance Monitoring	R	I	I	A	I	I	I
Internal Audit	C	I	I	I	I	I	I

RACI Legend:

• R = Responsible (does the work)
• A = Accountable (ultimately answerable)
• C = Consulted (provides input)
• I = Informed (kept up-to-date)

---

6.2 Role Descriptions

AI Act Program Manager

• Primary Responsibility: Executes classification process, maintains Classification Register, ensures compliance
• Key Activities:
  • Conducts prohibited practice and Annex III assessments
  • Applies classification decision tree
  • Maintains Classification Register
  • Coordinates annual reviews
  • Manages reclassification process
  • Reports metrics to AI Governance Committee
• Required Competencies: EU AI Act expertise, risk assessment, project management

AI System Owner

• Primary Responsibility: Accountable for accurate classification of their AI system
• Key Activities:
  • Requests classification
  • Provides complete AI system information
  • Participates in classification assessment
  • Approves classification decision
  • Monitors for reclassification triggers
  • Participates in annual reviews
• Required Competencies: Deep knowledge of AI system, business context, intended use

Product Director / CTO

• Primary Responsibility: Provides technical expertise, conducts Annex I assessments
• Key Activities:
  • Assesses if AI is safety component of Annex I products
  • Provides technical input on AI system capabilities
  • Reviews technical accuracy of classifications
• Required Competencies: Technical AI expertise, product safety knowledge

Legal & Compliance

• Primary Responsibility: Reviews borderline cases, ensures regulatory compliance
• Key Activities:
  • Reviews complex or borderline classification cases
  • Interprets EU AI Act requirements
  • Advises on legal implications of classifications
  • Manages regulatory authority interactions
• Required Competencies: EU AI Act legal expertise, regulatory compliance

AI Governance Committee

• Primary Responsibility: Approves high-risk classifications, provides oversight
• Key Activities:
  • Approves all HIGH-RISK classifications
  • Approves reclassifications
  • Reviews classification metrics quarterly
  • Escalation point for classification disputes
  • Oversight of classification compliance
• Authority: Final approval authority for high-risk classifications

Data Protection Officer (DPO)

• Primary Responsibility: Reviews privacy implications of classifications
• Key Activities:
  • Reviews classifications involving personal data processing
  • Advises on GDPR considerations
  • Identifies privacy risks in high-risk AI systems
• Required Competencies: GDPR expertise, privacy impact assessment

---

EXCEPTIONS

7.1 Exception Philosophy

Classification is a critical regulatory compliance activity. Exceptions are granted restrictively and only where compensating controls adequately mitigate risks.

---

7.2 Allowed Exceptions

The following exceptions may be granted with proper justification and approval:

Exception Type	Justification Required	Maximum Duration	Approval Authority	Compensating Controls
Delayed Classification (Pilot/Research)	AI system in early research phase; intended use unclear	6 months	AI Act Program Manager	Restrict to research environment; No deployment to production; Provisional "Assumed High-Risk" classification
Simplified Assessment (Clearly Minimal-Risk)	AI system obviously minimal-risk (e.g., simple spell checker)	Permanent	AI Act Program Manager	Document simplified rationale; Legal review if challenged; Annual re-confirmation
Extended Annual Review Timeline	Resource constraints prevent timely review	3 months extension	AI Governance Committee Chair	Prioritize high-risk systems; Interim monitoring; No new deployments until review complete

---

7.3 Prohibited Exceptions

The following exceptions cannot be granted under any circumstances:

❌ Skipping prohibited practice assessment - Mandatory for all AI systems, no exceptions
❌ Operating high-risk AI without proper classification - Regulatory violation, legal liability
❌ Ignoring reclassification triggers - Creates compliance gaps, regulatory risk
❌ Deploying prohibited AI practices - Explicitly forbidden by Article 5
❌ Incomplete documentation packages - Required for regulatory inspections

---

7.4 Exception Request Process

Step 1: Submit Exception Request

• Complete Exception Request Form (FORM-AI-EXCEPTION-001)
• Include business justification
• Propose compensating controls
• Specify duration requested
• Attach risk assessment

Step 2: Risk Assessment

• AI Act Program Manager assesses risk of granting exception
• Evaluates adequacy of compensating controls
• Documents residual risk

Step 3: Approval

• Route to appropriate approval authority based on exception type
• AI Act Program Manager approval: Minor exceptions (simplified assessment, short delays)
• AI Governance Committee Chair approval: Significant exceptions (extended timelines)
• AI Governance Committee approval: Critical exceptions

Step 4: Documentation and Monitoring

• Document exception in Exception Register
• Assign exception owner
• Set review date
• Monitor compensating controls
• Report exceptions quarterly to AI Governance Committee

Step 5: Exception Review and Closure

• Review exception at specified review date
• Assess if exception still needed
• Close exception when normal classification completed
• Document lessons learned

---

ENFORCEMENT

8.1 Non-Compliance Consequences

Violation	Severity	Consequence	Remediation Required
Operating AI without classification	Critical	Immediate suspension of AI system until classified	Complete classification within 5 business days; Root cause analysis; Corrective action plan
Incorrect classification (high-risk as lower)	Critical	Immediate reclassification; Compliance gap assessment; Halt deployment until gaps closed	Full compliance assessment; Gap remediation; Re-approval by AI Governance Committee
Prohibited practice deployed	Critical	Immediate shutdown; Incident investigation; Regulatory reporting (Article 73)	Decommission system; Legal review; Report to authorities; Disciplinary action
Missing annual review	High	Escalation to AI Governance Committee; Freeze on new features until reviewed	Complete review within 10 business days; Assess if reclassification needed
Late reclassification response	Medium	Written warning; Escalation to management	Complete reclassification immediately; Process improvement plan
Incomplete documentation package	Medium	Classification invalid until documentation complete	Complete documentation within 5 business days
Classification Register not updated	Low	Written warning	Update register immediately; Process correction

---

8.2 Escalation Procedures

Level 1: AI Act Program Manager

• Minor procedural violations
• Documentation deficiencies
• Timeline delays < 5 days
• Action: Written warning, corrective action required

Level 2: AI Governance Committee Chair

• Repeated violations
• High-risk misclassifications
• Missing annual reviews
• Action: Formal review, corrective action plan, management notification

Level 3: AI Governance Committee

• Prohibited practices identified
• Critical compliance failures
• Regulatory violations
• Action: Immediate AI system suspension, investigation, disciplinary action, regulatory reporting

Level 4: Executive Management + Legal

• Potential regulatory enforcement action
• Significant legal liability
• Reputational risk
• Action: Executive crisis management, legal strategy, regulatory engagement

---

8.3 Immediate Escalation Triggers

Escalate immediately to AI Governance Committee + Legal if:

• ⚠️ Prohibited practice (Article 5) suspected or identified
• ⚠️ High-risk AI system operating without proper classification or compliance
• ⚠️ Regulatory inquiry or inspection related to classification
• ⚠️ Classification dispute that cannot be resolved
• ⚠️ Potential serious incident (Article 73) related to classification error
• ⚠️ Media or public attention to AI system classification

---

8.4 Disciplinary Actions

Individuals responsible for classification violations may be subject to:

• Verbal or written warning
• Mandatory retraining
• Performance improvement plan
• Reassignment of responsibilities
• Suspension (with pay during investigation)
• Termination (for egregious violations, e.g., knowingly deploying prohibited practices)

Factors Considered:

• Intent (knowing violation vs. honest mistake)
• Severity of violation
• Impact (actual or potential)
• Cooperation with remediation
• Prior violation history

---

KEY PERFORMANCE INDICATORS (KPIs)

9.1 Classification KPIs

KPI ID	KPI Name	Definition	Target	Measurement Method	Frequency	Owner	Reporting To
KPI-CLS-001	Classification Completeness	% of AI systems with current, valid classification	100%	(# AI systems classified / # total AI systems) × 100	Monthly	AI Act Program Manager	AI Governance Committee
KPI-CLS-002	Classification Timeliness	Average business days from classification request to decision	< 10 days	Σ (classification date - request date) / # classifications	Monthly	AI Act Program Manager	Management
KPI-CLS-003	Annual Review Completion Rate	% of AI systems reviewed annually on schedule	100%	(# annual reviews completed on time / # annual reviews due) × 100	Quarterly	AI Act Program Manager	AI Governance Committee
KPI-CLS-004	Reclassification Responsiveness	% of triggered reclassifications completed within SLA	100%	(# reclassifications within SLA / # reclassification triggers) × 100	Monthly	AI Act Program Manager	AI Governance Committee
KPI-CLS-005	High-Risk Compliance Rate	% of high-risk AI systems with full compliance to required standards	100%	(# high-risk AI fully compliant / # high-risk AI systems) × 100	Quarterly	Compliance Officer	AI Governance Committee
KPI-CLS-006	Prohibited Practice Prevention	Number of prohibited practices in development or operation	0 (Zero tolerance)	Count of prohibited AI systems	Continuous	AI Governance Committee	Executive Management
KPI-CLS-007	Classification Accuracy	% of classifications validated as accurate in audits	≥ 95%	(# accurate classifications in sample / # classifications audited) × 100	Annually	Internal Audit	AI Governance Committee
KPI-CLS-008	Documentation Package Completeness	% of AI systems with complete documentation packages	100%	(# complete documentation packages / # AI systems) × 100	Monthly	Document Control	AI Act Program Manager
KPI-CLS-009	Authority Response Time	Average business days to respond to authority documentation requests	< 5 days	Σ (response date - request date) / # requests	Per request	Legal	Executive Management
KPI-CLS-010	Classification Register Accuracy	% of Classification Register fields complete and accurate	100%	(# complete/accurate records / # total records) × 100	Monthly	AI Act Program Manager	AI Governance Committee

---

9.2 KPI Dashboards and Reporting

Real-Time Dashboard (AI Act Program Manager access)

• Current classification completeness
• Classifications due this month
• Annual reviews due this quarter
• Reclassification triggers logged (pending action)
• Prohibited practice alerts

Monthly Management Report

• KPI-CLS-001, 002, 004, 008, 010
• Trend analysis (vs. previous month)
• Issues and risks
• Planned actions

Quarterly AI Governance Committee Report

• All KPIs
• High-risk AI system inventory and compliance status
• Annual review completion status
• Reclassifications completed this quarter
• Internal audit findings (if conducted)
• Exception register review
• Regulatory update impacts

Annual Executive Report

• Full-year KPI performance
• Classification accuracy validation (audit results)
• Compliance maturity assessment
• Strategic recommendations
• Regulatory outlook

---

9.3 KPI Thresholds and Alerts

KPI	Green (Good)	Yellow (Warning)	Red (Critical)	Alert Action
Classification Completeness	100%	95-99%	< 95%	Red: Immediate escalation to AI Governance Committee Chair
Classification Timeliness	< 10 days	10-15 days	> 15 days	Red: Escalate to AI Governance Committee; Review resource allocation
Annual Review Completion	100%	90-99%	< 90%	Yellow: Accelerate reviews; Red: Freeze new AI deployments until caught up
Reclassification Responsiveness	100%	90-99%	< 90%	Red: Escalate each overdue reclassification to AI Governance Committee
High-Risk Compliance	100%	95-99%	< 95%	Red: Halt high-risk AI deployments until compliant
Prohibited Practice Prevention	0	N/A	> 0	Red: Immediate escalation to Executive Management + Legal

---

TRAINING REQUIREMENTS

10.1 Training Program Overview

All personnel involved in AI classification must complete role-specific training to ensure competency in EU AI Act requirements and classification procedures.

---

10.2 Role-Based Training Requirements

Role	Training Course	Duration	Content	Frequency	Assessment Required
AI Act Program Manager	Classification Expert Training	16 hours	Deep dive into Article 5, Annex I, Annex III; Decision tree application; Edge cases; Legal considerations; Procedure mastery	Initial + annually	Yes - Written exam (≥90%) + Practical classification exercise
AI System Owners	Classification Overview	4 hours	Classification categories; Classification process; Responsibilities; How to request classification; Reclassification triggers	At onboarding + annually	Yes - Knowledge check (≥80%)
Product Directors / CTO	Annex I/III Assessment Training	6 hours	Annex I product safety requirements; Annex III use case analysis; Technical considerations; Safety component determination	Initial + annually	Yes - Practical assessment exercise
Legal & Compliance	EU AI Act Regulatory Training	8 hours	Full EU AI Act structure; Classification legal framework; Regulatory interpretation; Borderline case analysis; Authority engagement	Initial + annually	Yes - Written exam (≥90%)
All AI Development Staff	Prohibited Practices Awareness	2 hours	All 8 prohibited practices; Examples; How to spot prohibited practices; Escalation procedures	At onboarding + annually	Yes - Knowledge check (≥80%)
AI Governance Committee Members	AI Classification Governance	4 hours	Classification framework overview; Governance responsibilities; Approval criteria; Risk implications; Compliance oversight	At appointment + annually	Yes - Knowledge check (≥80%)
Document Control Staff	Classification Documentation Management	3 hours	Documentation package requirements; Document management procedures; Retention policies; Authority requests	Initial + as needed	Yes - Practical exercise

---

10.3 Training Content by Topic

Prohibited Practices Training (Article 5)

• Detailed explanation of all 8 prohibited practices
• Real-world examples and case studies
• How to identify prohibited practices in AI design
• Legal consequences of deploying prohibited practices
• Escalation procedures
• Recent regulatory guidance and enforcement actions

Annex I Training (Product Safety AI)

• Overview of 12 Annex I product categories
• Union harmonization legislation framework
• What constitutes a "safety component"
• Third-party conformity assessment requirements
• Integration with product safety processes
• Case studies of Annex I AI systems

Annex III Training (Standalone High-Risk AI)

• Detailed analysis of all 8 Annex III categories
• Specific use cases within each category
• How to assess intended purpose and deployment context
• Edge cases and borderline scenarios
• Recent regulatory interpretations
• Industry examples across all categories

Classification Decision Tree Training

• Step-by-step decision tree application
• How to use assessment forms
• Documentation requirements
• When to seek legal review
• Approval processes
• Common mistakes and how to avoid them

Classification Procedures Training

• PROC-AI-CLS-001: Classification Procedure
• PROC-AI-CLS-002: Review and Update Procedure
• PROC-AI-CLS-003: Documentation Management
• Hands-on exercises with classification tools
• Case studies and role-playing

---

10.4 Training Delivery Methods

Initial Training:

• Instructor-led classroom or virtual training
• Includes interactive exercises and case studies
• Hands-on practice with classification tools
• Group discussions of complex scenarios

Annual Refresher:

• E-learning modules for core content review
• Live update sessions for regulatory changes
• Case study reviews of recent classifications
• Knowledge assessment

On-the-Job Training:

• Mentoring for new AI Act Program Managers
• Job shadowing during classification process
• Supervised classifications for first 5 AI systems

Just-in-Time Training:

• Quick reference guides and job aids
• Video tutorials on specific topics
• Help desk support from experienced classifiers

---

10.5 Training Effectiveness Measurement

Assessment Methods:

• Written exams for knowledge retention
• Practical exercises for skill application
• On-the-job observations for competency validation
• Feedback surveys for training quality

Competency Validation:

• AI Act Program Managers: Must correctly classify 5 sample AI systems with 100% accuracy before independent classification
• AI System Owners: Must demonstrate understanding of classification categories and their responsibilities
• All staff: Must pass knowledge assessments with minimum required scores

Training Metrics:

Metric	Target	Frequency
Training completion rate	100%	Quarterly
Assessment pass rate (first attempt)	≥ 90%	Per training
Training effectiveness score (survey)	≥ 4.0/5.0	Per training
Time to competency (AI Act Program Managers)	< 30 days	Per person

---

10.6 Training Records

Records Maintained:

• Training attendance records
• Assessment scores
• Competency validations
• Refresher training completion
• Individual training transcripts

Retention: 10 years (to align with EU AI Act documentation retention)

Access: HR, AI Act Program Manager, Internal Audit, Competent Authorities (upon request)

---

DEFINITIONS

Term	Definition	Source
AI System	Machine-based system that, for explicit or implicit objectives, infers from input how to generate outputs (predictions, content, recommendations, decisions) that can influence physical or virtual environments	EU AI Act Article 3(1)
Prohibited AI Practice	AI system use explicitly forbidden by Article 5 of the EU AI Act due to unacceptable risks to safety, livelihoods, or fundamental rights	EU AI Act Article 5
High-Risk AI System	AI system listed in Annex III or serving as a safety component under Annex I, subject to mandatory requirements before market placement	EU AI Act Article 6, Annexes I & III
Limited-Risk AI	AI system with specific transparency obligations under Article 50 (e.g., chatbots, deepfakes, emotion recognition, biometric categorization not covered by prohibited or high-risk categories)	EU AI Act Article 50
Minimal-Risk AI	AI system not classified as prohibited, high-risk, or limited-risk; no specific EU AI Act obligations but voluntary codes of conduct encouraged	EU AI Act (default category)
Provider	Natural or legal person, public authority, agency, or other body that develops an AI system or has an AI system developed with a view to placing it on the market or putting it into service under its own name or trademark, whether for payment or free of charge	EU AI Act Article 3(3)
Deployer	Natural or legal person, public authority, agency, or other body using an AI system under its authority, except where the AI system is used in the course of a personal non-professional activity	EU AI Act Article 3(4)
Safety Component	Component of a product or system which fulfills a safety function for that product or system or the failure or malfunctioning of which endangers the health and safety of persons or property	EU AI Act Article 3(14)
Substantial Modification	Change to an AI system after its placing on the market or putting into service which affects the compliance of the AI system with the requirements of the EU AI Act or results in a modification to the intended purpose for which the AI system has been assessed	EU AI Act Article 3(23)
Intended Purpose	Use for which an AI system is intended by the provider, including the specific context and conditions of use, as specified in the information supplied by the provider in the instructions for use, promotional or sales materials and statements, as well as in the technical documentation	EU AI Act Article 3(12)
Reasonably Foreseeable Misuse	Use of an AI system in a way that is not in accordance with its intended purpose, but which may result from reasonably foreseeable human behavior or interaction with other systems	EU AI Act Article 3(13)
Classification	Process of determining which EU AI Act risk category (prohibited, high-risk, limited-risk, or minimal-risk) applies to a specific AI system based on its characteristics, intended purpose, and deployment context	This Standard
Classification Register	Centralized database of all AI system classifications maintained by the organization, including classification decisions, rationale, dates, approvers, and current status	This Standard
Reclassification	Process of reassessing and potentially changing an AI system's classification due to substantial modifications, intended purpose changes, deployment context changes, or regulatory updates	This Standard
Annex I Product	Product covered by Union harmonization legislation listed in Annex I of the EU AI Act (e.g., machinery, medical devices, toys) where AI may serve as a safety component	EU AI Act Annex I
Annex III Use Case	Specific high-risk use case listed in Annex III of the EU AI Act across 8 categories: biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, and justice/democracy	EU AI Act Annex III
Biometric Data	Personal data resulting from specific technical processing relating to the physical, physiological, or behavioral characteristics of a natural person, which allow or confirm the unique identification of that natural person, such as facial images or dactyloscopic (fingerprint) data	EU AI Act Article 3(34) / GDPR Article 4(14)
Sensitive Attributes	Personal data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, trade union membership, genetic data, biometric data (for unique identification), health data, or data concerning sex life or sexual orientation	GDPR Article 9(1)
Conformity Assessment	Process demonstrating whether the requirements of the EU AI Act relating to a high-risk AI system have been fulfilled	EU AI Act Article 3(20)
Placing on the Market	First making available of an AI system on the Union market	EU AI Act Article 3(9)
Putting into Service	Supply of an AI system for first use directly to the deployer or for own use in the Union for its intended purpose	EU AI Act Article 3(11)
CE Marking	Marking by which a provider indicates that an AI system is in conformity with the requirements set out in the EU AI Act and other applicable Union legislation harmonising the conditions for the marketing of products	EU AI Act Article 3(24)

---

LINK WITH AI ACT AND ISO42001

12.1 EU AI Act Regulatory Mapping

This standard implements the following EU AI Act requirements:

EU AI Act Provision	Article/Annex	Requirement Summary	Implemented By (Controls)
AI System Definition	Article 3(1)	Defines what constitutes an AI system subject to the regulation	Scope (Section 2.2)
Prohibited AI Practices	Article 5	Lists 8 AI practices that are prohibited due to unacceptable risk	CLS-001
Classification Rules for High-Risk AI	Article 6	Defines how to identify high-risk AI systems	CLS-002, CLS-003, CLS-004
Annex I: Product Safety AI	Annex I	Lists 12 product categories where AI safety components are high-risk	CLS-002
Annex III: Standalone High-Risk Use Cases	Annex III	Lists 8 categories of high-risk standalone AI systems	CLS-003
Transparency Obligations	Article 50	Requires disclosure for certain AI systems (chatbots, deepfakes, etc.)	CLS-004 (decision tree)
Provider Obligations	Articles 16, 17, 25	Requires providers to ensure AI systems are classified and compliant	CLS-005, CLS-006, CLS-007
Documentation Requirements	Article 11, Annex IV	Requires technical documentation (classification is prerequisite)	CLS-009
Record Keeping	Article 12, Article 18	Requires 10-year retention of documentation	CLS-009
Conformity Assessment	Article 43	Requires conformity assessment for high-risk AI (classification determines procedure)	CLS-004 (triggers STD-AI-010)

---

12.2 ISO/IEC 42001:2023 Alignment

This standard aligns with ISO/IEC 42001:2023 (Artificial Intelligence Management System) as follows:

ISO 42001 Clause	Requirement	Implementation in This Standard
Clause 4.1: Understanding the organization and its context	Understand external issues affecting AI management system (including legal/regulatory requirements)	Section 2 (Scope and Applicability); Section 12.1 (EU AI Act mapping)
Clause 4.2: Understanding needs and expectations of interested parties	Identify interested parties and their requirements	Roles & Responsibilities (Section 6); Stakeholder communication through classification
Clause 5.2: AI policy	Establish AI policy including regulatory compliance commitment	Links to parent POL-AI-001; Classification as first step in compliance
Clause 6.1: Actions to address risks and opportunities	Risk-based approach to AI management	Classification determines risk level and compliance obligations
Clause 6.1.2: AI system impact assessment	Assess AI systems for risks to individuals, groups, society, environment	Prohibited practice assessment (CLS-001); High-risk assessment (CLS-002, CLS-003)
Clause 6.2: AI objectives	Establish measurable AI objectives	KPIs (Section 9)
Clause 7.2: Competence	Ensure persons have appropriate competence	Training Requirements (Section 10)
Clause 7.5: Documented information	Maintain documented information	Classification documentation packages (CLS-009)
Clause 8.2: AI system risk assessment	Establish process for AI risk assessment throughout lifecycle	Classification as initial risk assessment; Links to STD-AI-002 (Risk Management)
Clause 9.1: Monitoring, measurement, analysis, and evaluation	Monitor and measure AI management system performance	Compliance monitoring (Section 5); KPIs (Section 9)
Clause 9.2: Internal audit	Conduct internal audits	Internal audit requirements (Section 5.2)
Clause 9.3: Management review	Conduct management reviews	AI Governance Committee quarterly reviews
Clause 10.1: Nonconformity and corrective action	Address nonconformities	Enforcement (Section 8); Reclassification (CLS-007, CLS-008)
Clause 10.2: Continual improvement	Continually improve AI management system	Annual reviews (CLS-006); KPI tracking; Procedure improvements

---

12.3 Relationship to Other Standards

This classification standard is the foundational standard that determines which other AI Act standards apply:

Classification Result	Triggered Standards (Mandatory)	Rationale
PROHIBITED	None (system must not be developed/deployed)	Article 5 prohibition
HIGH-RISK (Annex I or III)	All of the following:	Articles 8-15, 17-18, 43, 49, 72-73
	• STD-AI-002: Risk Management	Article 9
	• STD-AI-003: Data Governance	Article 10
	• STD-AI-004: Technical Documentation	Article 11
	• STD-AI-005: Logging and Record Keeping	Article 12
	• STD-AI-006: Transparency	Article 13
	• STD-AI-007: Human Oversight	Article 14
	• STD-AI-008: Accuracy, Robustness, Security	Article 15
	• STD-AI-009: Quality Management	Article 17
	• STD-AI-010: Conformity Assessment	Article 43
	• STD-AI-011: Registration	Article 49
	• STD-AI-012: Post-Market Monitoring	Article 72
	• STD-AI-013: Incident Management	Article 73
	• STD-AI-014: AI Literacy and Training	Article 4
LIMITED-RISK	• STD-AI-006: Transparency (Article 50 obligations only)	Article 50
MINIMAL-RISK	None (voluntary best practices recommended)	No specific EU AI Act obligations

Classification Decision Impact:

• Classification determines regulatory burden (none to comprehensive)
• Classification determines time to market (minimal-risk: fast; high-risk: months for conformity assessment)
• Classification determines ongoing compliance costs (minimal-risk: low; high-risk: significant)
• Classification determines regulatory oversight (minimal-risk: none; high-risk: authority supervision)

---

12.4 References and Related Documents

EU AI Act (Regulation (EU) 2024/1689):

• Full text: [Official Journal of the European Union]
• Article 3: Definitions
• Article 5: Prohibited AI Practices
• Article 6: Classification Rules for High-Risk AI Systems
• Annex I: List of Union Harmonization Legislation
• Annex III: High-Risk AI Systems
• Article 50: Transparency Obligations for Certain AI Systems

ISO/IEC Standards:

• ISO/IEC 42001:2023: Information technology — Artificial intelligence — Management system
• ISO/IEC 23894:2023: Information technology — Artificial intelligence — Guidance on risk management
• ISO/IEC 5338:2023: Information technology — Artificial intelligence — AI system life cycle processes

Internal Documents:

• POL-AI-001: Artificial Intelligence Policy (parent policy)
• STD-AI-002 through STD-AI-014: Related AI Act compliance standards
• PROC-AI-CLS-001, -002, -003: Classification procedures

External Guidance:

• European Commission: AI Act Implementation Guidance
• European AI Office: Classification Guidelines
• National Competent Authorities: Jurisdiction-specific guidance

---

APPROVAL AND AUTHORIZATION

Role	Name	Title	Signature	Date
Prepared By	Sarah Johnson	AI Act Program Manager	_________________	________
Reviewed By	Michael Brown	Chief Legal Officer	_________________	________
Reviewed By	David Lee	Chief Technology Officer	_________________	________
Reviewed By	Jane Doe	Chief Strategy & Risk Officer	_________________	________
Approved By	Jane Doe	AI Governance Committee Chair	_________________	________

Effective Date: 2025-08-01
Next Review Date: 2026-08-01
Review Frequency: Annually or upon regulatory change

---

END OF STANDARD STD-AI-001

---

This standard is a living document. Feedback and improvement suggestions should be directed to the AI Act Program Manager.