Markdown Editor

AI Act Article 6(3) Not-High-Risk Assessment Template

Provider record that an Annex III AI system is not high-risk under Art. 6(3): profiling check, condition relied on, risk reasoning, Art. 49(2) registration

ШІ та машинне навчанняСтандартEU AI Act Art. 6(3)-(4), Art. 49(2) (as amended by 2026/1744)acthighriskexemptionassessment

Як використовувати: Use it when a provider believes an Annex III AI system is not high-risk under Article 6(3). Fill the profiling check first, since profiling always means high-risk, then one condition with evidence. The main pitfall is relying on all four conditions loosely instead of one well-evidenced one.

Перегляд

Article 6(3) Not-High-Risk Assessment: [AI system] [version]

Documented assessment by a provider that an AI system falling within an Annex III area is not high-risk under Article 6(3) of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, the Digital Omnibus on AI), with the Article 6(4) documentation duty and the Article 49(2) registration in the EU database. It records the Annex III area, the profiling check, the condition relied on, the significant-risk reasoning, the evidence, and the triggers for reassessment. Replace every [bracketed] value and every row marked Example. Under the amended Article 113, Chapter III Sections 1 to 3 apply from 2 December 2027 for Annex III systems; the Commission published draft classification guidelines on 19 May 2026. Check the consolidated text and the final guidelines. This is a documentation format, not legal advice: check the current official text and a qualified professional.

Document control

FieldValue
Document ID[EXA-001]
Provider[Legal name, address]
AI system and version[Name, version]
Intended purpose (as stated in documentation)[One sentence]
AI system inventory ID[AIS-ID]
Guidance relied on[Commission guidelines under Article 6(5), version and date]
Version[0.1]
Status[Draft / Approved / Registered / Superseded]
Owner[@product-owner]
Assessors[@regulatory-lead], [@legal], [@ml-lead], [@fundamental-rights-expert]
Approved before placing on market[Yes], YYYY-MM-DD
EU database registration reference[Reference]
Last updatedYYYY-MM-DD

Approvals

RoleNameDecisionDate
Regulatory lead[@name][Approve / Reject: treat as high-risk]YYYY-MM-DD
Legal counsel[@name][Decision]YYYY-MM-DD
Accountable executive[@name][Decision]YYYY-MM-DD

Revision history

VersionDateAuthorChange summary
0.1YYYY-MM-DD[@name]First draft
1.0YYYY-MM-DD[@name]Example: Approved; registered under Article 49(2)
1.1YYYY-MM-DD[@name]Example: Reassessed after proposed rank-based slot feature (section 7)

Contents

1. System description

ItemValue
FunctionExample: CRV Schedule Assist by Tarnwick Analytics proposes interview time slots for candidates a recruiter has already chosen to invite
InputsExample: Recruiter and interviewer calendars, candidate-stated availability, room bookings
OutputsExample: Three proposed slots per candidate, sent after recruiter confirmation
Who uses the outputExample: Recruiters and candidates
Decisions it feedsExample: None about suitability; only the time of an interview already decided by a human
Data not usedExample: CV content, application answers, scores from CRV Shortlist or any other ranking

2. Annex III area

QuestionAnswer
Annex III pointExample: Point 4(a), AI systems intended for recruitment or selection of natural persons
Reason it falls within the areaExample: The system is part of the recruitment process and is marketed to recruiters
Annex I product?[No]
Same provider's other systems in the processExample: CRV Shortlist 2.4, which is treated as high-risk and documented separately

3. Profiling check

QuestionAnswerReasoning and evidence
Does the system process personal data?[Yes / No]Example: Yes, names, contact details and availability
Does it evaluate personal aspects of a person?[Yes / No]Example: No. It matches free time; it does not assess the candidate
Does it analyse or predict behaviour, reliability, preferences or performance?[Yes / No]Example: No. Slot order is by earliest date, not by predicted attendance
Could configuration options turn it into profiling?[Yes / No]Example: No; no option to weight candidates. Section 7 lists features that would change this
Profiling result[Profiling / No profiling]If profiling: stop, the system is high-risk

4. Condition relied on (Article 6(3)(a) to (d))

RefCondition (summarised)Relied on?Reasoning
(a)The system performs a narrow procedural task[Yes / No]Example: Yes. It converts a decision already taken (invite) into a calendar slot; it has no effect on whether the person is invited or selected
(b)The system improves the result of a human activity already completed[Yes / No]Example: No
(c)The system detects decision patterns or deviations from earlier patterns, is not meant to replace or influence the completed human assessment, and is subject to proper human review[Yes / No]Example: No
(d)The system performs a preparatory task to an assessment relevant to an Annex III use case[Yes / No]Example: No. Scheduling follows the assessment, it does not prepare it

Evidence:

IDEvidenceLocation
EV-01 (Example)Functional specification showing inputs and outputsSpec SA-SPEC-3
EV-02 (Example)Data flow diagram showing no link to ranking scoresArchitecture record AR-11
EV-03 (Example)Test showing slot proposals do not vary with application contentTest report TR-45

5. Significant risk and material influence

IDPossible harmRights concernedAssessmentMitigation
H-01 (Example)Candidates with caring duties get fewer convenient slotsNon-discriminationLow: candidates state their own availability and may request other times"None of these times work" option always shown
H-02 (Example)Missed interview due to scheduling error affects outcomeNon-discrimination, good administrationLow: the error affects timing, recruiter can rescheduleRescheduling link in every message
H-03 (Example)Accessibility needs not capturedIntegration of persons with disabilitiesLow with mitigationFree-text adjustment request routed to recruiter

Material influence conclusion: [Example: The system does not materially influence whether a candidate is selected. It acts only after the human invitation decision and does not use any information about the candidate's suitability.]

6. Registration and documentation duties

DutyStatusEvidence
Assessment documented before placing on the market or putting into service (Art. 6(4))[Done], YYYY-MM-DDThis document, version [x]
Provider and system registered in the EU database (Art. 49(2))[Done / Pending], YYYY-MM-DD[Reference]
Annex VIII Section B information submitted[Done]Provider and contact details, trade name and reference, intended purpose, the Article 6(3) condition relied on, status
Documentation provided to national competent authority on request (Art. 6(4))[Not requested / Provided YYYY-MM-DD][Reference]
Other duties still apply[Checked]Article 50 transparency where relevant, Article 4 literacy measures, GDPR

7. Reassessment triggers

TriggerWhy it mattersAction
Example: Proposed feature to offer earlier slots to higher-ranked candidatesUses suitability scores: influences outcome and may be profilingNew assessment; likely high-risk
Example: Predicting no-show likelihood to overbook slotsPredicts behaviour: profilingTreat as high-risk
New intended purpose or customer segmentMay change Annex III areaReassess
Final Commission guidelines published or changedInterpretation may narrow the conditionReassess within [30 days]
Market surveillance authority query (Art. 80)Possible reclassificationEscalate to legal
Scheduled reviewDrift between documentation and productEvery [12 months]

8. Conclusion

ItemValue
Outcome[Not high-risk under Article 6(3) / High-risk]
Condition relied onExample: Article 6(3)(a), narrow procedural task
ProfilingExample: None
Conditions attachedExample: No link to ranking scores; features in section 7 need a new assessment
Decision by[@accountable-executive], YYYY-MM-DD

If the outcome is high-risk, continue with ai-data/ai-act-risk-management-system, ai-data/ai-act-technical-documentation, ai-data/ai-act-instructions-for-use and ai-data/ai-act-declaration-of-conformity.

9. Glossary and references

TermMeaning
Annex IIIList of high-risk use case areas in the AI Act, such as employment, education and essential services
ProfilingAutomated processing of personal data to evaluate personal aspects of a person (GDPR Article 4(4))
Narrow procedural taskA limited task that does not shape the substance of a decision
EU databaseThe database for high-risk AI systems listed in Annex III (Article 71)
Market surveillance authorityNational authority that checks AI systems on the market

References (check the current consolidated text and guidelines):

Standard and licence note

Based on Regulation (EU) 2024/1689 (the Artificial Intelligence Act) as amended by Regulation (EU) 2026/1744 (Digital Omnibus on AI), Article 6(3) and (4), Article 49(2), Article 80 and Annex VIII Section B, as published in the Official Journal of the European Union; EU legislative texts are free to reuse. The conditions are summarised in original wording and no text of the Regulation is reproduced. Check the consolidated text on EUR-Lex and the final Commission guidelines under Article 6(5) before relying on an assessment. All organisations, products and figures in examples are fictional. This is a documentation format, not legal advice. Template licence: MIT.

Схожі шаблони