If our AI falls in an Annex III area, is it automatically high-risk?

No - not automatically. Article 6(3) provides a filter: if the system does not pose a significant risk of harm to health, safety or fundamental rights, it is not classified high-risk even if it falls in an Annex III area. However, the filter does not apply if the system profiles people, and you must document your reasoning.

Does the Annex I route require a notified body?

It depends on the underlying product-safety legislation. For many Annex I product categories (e.g. certain medical devices, vehicles), the sectoral law already requires third-party conformity assessment via a notified body, and AI components follow the same route. For some Annex I categories the provider self-assessment is sufficient.

When do high-risk obligations apply?

Under current law (as enacted), most high-risk obligations apply from 2 August 2026. For Annex I safety-component AI, the deadline is 2 August 2027. The Digital Omnibus provisional agreement (not yet law, 9 June 2026) would push the stand-alone Annex III deadline to 2 December 2027 and the Annex I deadline to 2 August 2028.

Is an AI system used internally (not sold) subject to the high-risk rules?

Yes, if a public authority or other organisation deploys a high-risk AI system under its own authority in a professional context, it is a "deployer" and carries the Article 26 deployer obligations. If it also puts the system into service under its own name, it may be reclassified as a provider (Article 25) with the full provider obligations.

Does emotion recognition always trigger the biometrics area in Annex III?

Emotion recognition falls in Annex III area 1 (biometrics) where it is not already prohibited under Article 5 (i.e. it is used outside workplaces and educational institutions, or for medical/safety reasons that exempt it from the Article 5 ban). Even then, the Article 6(3) filter may still apply if the specific deployment does not pose significant risk.

EU AI Act topic guide

High-risk AI systems: Annex III, Annex I and the Article 6(3) filter

Not all AI is created equal, and the AI Act's most demanding obligations fall on "high-risk AI systems." There are two independent routes to this classification. Understanding which route applies - and whether the Article 6(3) filter can bring a system out of scope - is the most consequential compliance question for most organisations.

Reviewed by the AI Act Navigator team · Last updated 9 June 2026

TL;DR

Route 1 (Annex I): AI that is a safety component of, or itself a regulated product under, EU harmonisation law (machinery, medical devices, toys, motor vehicles, etc.) and required to undergo third-party conformity assessment.
Route 2 (Annex III): Stand-alone AI in one of 8 listed use-case areas - biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration/asylum, justice/democracy.
Article 6(3) filter: an Annex III system is NOT high-risk if it does not pose significant risk of harm - e.g. narrow procedural tasks, pattern detection without replacing human judgement. Exception: does not apply if the system profiles people.
Obligations apply from 2 August 2026 ([OMNIBUS - PROPOSED] deferred to 2 December 2027 for stand-alone Annex III if the Digital Omnibus is adopted).

Classify your AI system Obligations guide

Scope

What this covers

Annex I route: AI that is a safety component of products covered by the CE-marking directives/regulations listed in Annex I, including the Machinery Regulation, Medical Devices Regulation, In Vitro Diagnostic Medical Devices Regulation, Radio Equipment Directive, General Product Safety Regulation, Automotive type-approval, Aviation, Toys, Lifts, Pressure equipment, and Personal Protective Equipment.
Annex III area 1 - Biometrics: remote biometric identification, biometric categorisation, and emotion recognition not covered by Article 5 prohibitions.
Annex III area 2 - Critical infrastructure: AI used as safety components in management or operation of road traffic, water, gas, heating, electricity, and digital infrastructure.
Annex III area 3 - Education & vocational training: AI that determines access or assignment to educational/vocational institutions, assesses learning outcomes, evaluates learning level, or monitors/proctors exams.
Annex III area 4 - Employment & workers management: recruitment and candidate screening, promotion and termination decisions, task allocation and monitoring of performance or behaviour.
Annex III area 5 - Access to essential private & public services: credit-scoring and creditworthiness evaluation (excluding fraud detection); eligibility for public benefits; life and health insurance risk assessment and pricing; emergency service dispatch and triage.
Annex III area 6 - Law enforcement: risk assessment of offending or re-offending, polygraphs, evidence-reliability evaluation, criminal profiling in investigations.
Annex III area 7 - Migration, asylum & border control: polygraphs, risk assessments, examination of asylum/visa/residence applications, detection or recognition of persons.
Annex III area 8 - Administration of justice & democratic processes: AI assisting judicial authorities in researching, interpreting or applying the law; AI influencing election or referendum outcomes or voting behaviour.

Article 6(3) filter: even if an AI system falls in an Annex III area, it is NOT classified as high-risk if it does not pose a significant risk of harm to health, safety or fundamental rights. Indicators of non-significant risk: the system performs a narrow procedural task; it improves a prior human activity; it only detects decision-making patterns without replacing human judgement; it performs preparatory tasks. Important: the filter does not apply where the system profiles natural persons.

Source: Regulation (EU) 2024/1689 (EUR-Lex)

Compliance challenges

Key compliance challenges

Determining which Annex III area applies is fact-sensitive - the same AI capability (e.g. an ML classifier) can fall inside or outside scope depending on its intended purpose and deployment context.
The Article 6(3) filter requires a documented assessment: organisations must be able to justify why their Annex III system does not pose significant risk of harm.
Annex I coverage is wide: any AI embedded in CE-marked products (machinery, medical devices, vehicles) as a safety component likely triggers the Annex I route, which may require involving a notified body.
Timeline uncertainty: the original 2 August 2026 deadline for high-risk obligations may move to 2 December 2027 for Annex III systems under the Digital Omnibus, but this is not yet law as of 9 June 2026.

The EU AI Act applies a risk-based approach: obligations scale with the level of risk posed. AI Act high-level summary

What to do

Inventory every AI system in your portfolio: document its intended purpose, deployment context and the population of users or affected persons.
Apply the Annex I test: does the AI operate as a safety component of, or is it itself a product governed by, any of the Annex I directives/regulations?
Apply the Annex III test: does the AI fall in one of the 8 listed use-case areas given its intended purpose?
Apply the Article 6(3) filter to Annex III candidates: can you substantiate that the system does not pose significant risk of harm? Document the assessment.
For confirmed high-risk systems: build the Chapter III compliance programme - risk management system, data governance, technical documentation, logging, human oversight, accuracy/robustness, quality management, conformity assessment, EU declaration of conformity, CE marking, registration, post-market monitoring.

For the full obligations breakdown, see the AI Act obligations guide, and for role-specific duties see the provider vs deployer guide.

FAQ

High-risk AI systems: common questions

If our AI falls in an Annex III area, is it automatically high-risk?: No - not automatically. Article 6(3) provides a filter: if the system does not pose a significant risk of harm to health, safety or fundamental rights, it is not classified high-risk even if it falls in an Annex III area. However, the filter does not apply if the system profiles people, and you must document your reasoning.
Does the Annex I route require a notified body?: It depends on the underlying product-safety legislation. For many Annex I product categories (e.g. certain medical devices, vehicles), the sectoral law already requires third-party conformity assessment via a notified body, and AI components follow the same route. For some Annex I categories the provider self-assessment is sufficient.
When do high-risk obligations apply?: Under current law (as enacted), most high-risk obligations apply from 2 August 2026. For Annex I safety-component AI, the deadline is 2 August 2027. The Digital Omnibus provisional agreement (not yet law, 9 June 2026) would push the stand-alone Annex III deadline to 2 December 2027 and the Annex I deadline to 2 August 2028.
Is an AI system used internally (not sold) subject to the high-risk rules?: Yes, if a public authority or other organisation deploys a high-risk AI system under its own authority in a professional context, it is a "deployer" and carries the Article 26 deployer obligations. If it also puts the system into service under its own name, it may be reclassified as a provider (Article 25) with the full provider obligations.
Does emotion recognition always trigger the biometrics area in Annex III?: Emotion recognition falls in Annex III area 1 (biometrics) where it is not already prohibited under Article 5 (i.e. it is used outside workplaces and educational institutions, or for medical/safety reasons that exempt it from the Article 5 ban). Even then, the Article 6(3) filter may still apply if the specific deployment does not pose significant risk.

Get AI Act-ready

Use the risk classifier to find your system's tier, then explore the obligations and checklist for your role.

Classify your AI system

This is guidance, not legal advice

This is guidance to help you understand how the EU AI Act applies to high-risk ai systems, not legal advice. For decisions specific to your organisation, confirm with the official sources we link or a qualified legal adviser.

Sources

[1]Regulation (EU) 2024/1689 (EU AI Act) - EUR-Lexretrieved 9 Jun 2026
[2]European Commission: AI regulatory frameworkretrieved 9 Jun 2026
[3]AI Act Explorer: high-level summaryretrieved 9 Jun 2026
[4]AI Act implementation timelineretrieved 9 Jun 2026
[5]Council of the EU: Digital Omnibus provisional agreement, 7 May 2026retrieved 9 Jun 2026

The AI Act Brief

Subscribe to The AI Act Brief

We watch Brussels so you don't. Plain-English EU AI Act updates, free.

No spam. Unsubscribe anytime.

categoryWhat this covers

priority_highKey compliance challenges

checklistWhat to do

helpHigh-risk AI systems: common questions