Checklist · AI Act Chapter III
Status: 4 October 2026
High-Risk AI Checklist: All Obligations under the EU AI Act
Your AI system is high-risk – now what? This checklist turns Chapter III of the AI Act into tasks you can tick off: separately for providers and deployers, with article, evidence and deadlines. Interactive in the browser or as a free Excel file for your project team.
- 17 obligations for providers
- 14 points for deployers
- Evidence and retention periods
- Deadlines after the Digital Omnibus
High-risk checklist
Applicant screening · Annex III point 4
- Role checked: deployer, not providerArt. 25
- Overseers appointed and enabledArt. 26(2)
- Works council and staff informedArt. 26(7)
- Logs kept for at least six monthsArt. 26(6)
- Affected persons informedArt. 26(11)
Sample data · deployer obligations apply from 2 December 2027
In short
The AI Act assigns high-risk obligations by role. Providers need, among other things, risk management, data governance, technical documentation, quality management, a conformity assessment, CE marking and registration. Deployers must use the system as instructed, ensure human oversight, keep logs and inform staff and affected persons. For Annex III, the obligations apply from 2 December 2027.
Context
What the checklist covers – and who it applies to
The obligations in Chapter III of the AI Act (Regulation (EU) 2024/1689) only apply to high-risk AI systems. Under Art. 6, a system is high-risk in two ways: as a safety component or product covered by the harmonisation legislation in Annex I that requires a third-party conformity assessment (para. 1), or as a use case listed in Annex III, for example in employment, creditworthiness assessment or education (para. 2). Annex III systems can be exempt if they do not pose a significant risk (para. 3) – but profiling of natural persons is always excluded from that exemption.
This page starts where classification ends. Which areas and use cases Annex III covers and how the exemption is tested is explained in detail on our Annex III page. Here the question is what comes next: what exactly has to be done, documented and evidenced?
The obligations depend on your role. A provider develops an AI system or has it developed and places it on the market or puts it into service under its own name (Art. 3(3)). A deployer uses an AI system under its own authority, except for purely personal activities (Art. 3(4)). Most obligations sit with the provider: Art. 16 bundles them and refers to the requirements for the system in Art. 8 to 15 as well as quality management, documentation, conformity assessment and registration. Deployers’ obligations are mainly in Art. 26, supplemented by the fundamental rights impact assessment in Art. 27 for certain deployers.
Most mid-sized companies are deployers: they buy an HR tool with AI pre-screening, a scoring solution or control-room software. Their list is shorter but not trivial – human oversight by competent people, keeping the logs and informing staff and affected persons in particular need preparation. Software companies that build AI features into such products, on the other hand, are providers and carry the full list.
On timing, the Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026) postponed the high-risk obligations: for Annex III to 2 December 2027, for Annex I to 2 August 2028. The prohibitions under Art. 5 and the duty to take AI literacy measures under Art. 4 already apply. For systems placed on the market before the cut-off date, the obligations generally only apply after a significant change; high-risk AI used by public authorities must comply by 2 August 2030 (Art. 111(2)).
Infringements of providers’ obligations under Art. 16 and deployers’ obligations under Art. 26 can be fined up to €15 million or 3% of worldwide annual turnover; for SMEs the lower amount applies (Art. 99(4) and (6)). The checklist below is built as a working list: one row per task, one article, one piece of evidence.
- obligations for high-risk AI under Annex III apply (Digital Omnibus)
- 2 Dec 2027
- obligations for high-risk AI under Annex III apply (Digital Omnibus)
- or 3% of turnover: fine ceiling for infringements of Art. 16 and 26 (Art. 99(4))
- €15m
- or 3% of turnover: fine ceiling for infringements of Art. 16 and 26 (Art. 99(4))
- retention of technical documentation and the declaration of conformity by the provider
- 10 years
- retention of technical documentation and the declaration of conformity by the provider
Not yet sure whether your system is high-risk?
Settle that first. The checklist assumes the classification is done.
Interactive checklist
Tick off the obligations: checklist for providers and deployers
Choose your role and the characteristics that apply to you. The list then shows only the obligations relevant to you – with article, task and possible evidence. On the right you see progress and the open points, which you can copy as a task list. Your input stays in your browser.
Your status
0 of 11 done
Nothing ticked yet. Start on the left with the first point.
Simplified overview, not legal advice. Points marked “Recommendation” are not a legal obligation.
Evidence
Which documents and evidence high-risk AI requires
In the end, what counts is what you can show an authority. The table lists the documents that result from the obligations – who produces them, what they are based on, what they must contain at a minimum and how long to keep them.
Scroll the table sideways →
| Document | Who | Reference | Core content | Retention |
|---|---|---|---|---|
| Risk management documentation | Provider | Art. 9 | Identified risks, evaluation, measures, residual risks, tests; continuously updated. | Part of the technical documentation |
| Technical documentation | Provider | Art. 11, Annex IV | At least the elements in Annex IV; SMEs and small mid-caps simplified via the Commission form. | 10 years (Art. 18) |
| Instructions for use | Provider → deployer | Art. 13(3) | Intended purpose, accuracy and metrics, performance limits, oversight measures, maintenance, logs. | With the system |
| Quality management system | Provider | Art. 17 | Written policies and procedures from change management to accountability; proportionate to size. | 10 years (Art. 18) |
| EU declaration of conformity | Provider | Art. 47 | Declaration per system, written, machine-readable, physically or electronically signed. | 10 years (Art. 47(1)) |
| Registration in the EU database | Provider; authorities as deployers | Art. 49 | Details on provider and system; also when relying on the Art. 6(3) exemption. | Keep up to date |
| Post-market monitoring plan | Provider | Art. 72 | How performance data is collected and analysed over the lifetime. | Part of the technical documentation |
| Logs | Provider and deployer | Art. 12, 19, 26(6) | Automatically generated records, where under own control. | at least 6 months |
| Fundamental rights impact assessment | Certain deployers | Art. 27 | Processes, period, affected groups, risks, oversight, remedies; notification to the authority. | Update on changes |
| Data protection impact assessment | Deployer (as controller) | Art. 26(9), Art. 35 GDPR | Where a high risk is likely; include the provider’s Art. 13 information. | Update on changes |
| Information to staff | Deployer as employer | Art. 26(7) | Before use at the workplace, to workers’ representatives and affected workers. | Recommendation: with the AI inventory |
| Serious incident reports | Provider; deployer to provider | Art. 73, 26(5) | Incident, link to the system, measures; an incomplete initial report is possible. | Recommendation: incident register |
Simplified overview. Where the law sets no period, retention is marked as a recommendation. Data protection deletion duties remain unaffected.
Deadlines
Key dates, reporting periods and retention at a glance
The most important dates for high-risk AI after the Digital Omnibus – from application to reporting a serious incident.
Scroll the table sideways →
| Topic | Deadline | Reference |
|---|---|---|
| High-risk obligations for Annex III apply | 2 December 2027 (before: 2 August 2026) | Regulation (EU) 2026/1744 |
| High-risk obligations for Annex I apply | 2 August 2028 (before: 2 August 2027) | Regulation (EU) 2026/1744 |
| High-risk AI for public authorities already on the market | by 2 August 2030 | Art. 111(2) |
| Reporting a serious incident – general rule | no later than 15 days after becoming aware | Art. 73(2) |
| Reporting a widespread infringement or serious disruption of critical infrastructure | no later than 2 days after becoming aware | Art. 73(3) |
| Reporting in the event of a death | no later than 10 days after becoming aware | Art. 73(4) |
| Technical documentation, QMS, declaration of conformity | 10 years from placing on the market | Art. 18(1), Art. 47(1) |
| Automatically generated logs | at least 6 months | Art. 19(1), Art. 26(6) |
Role change
When deployers become providers
Art. 25(1) names three cases in which a deployer, distributor or importer is considered a provider – with all the provider obligations in this checklist. The original provider must then cooperate closely and make the necessary information available.
Own name or trade mark
You put your name or trade mark on a high-risk system already placed on the market – for example as a white-label solution for your customers.
Art. 25(1)(a)
Substantial modification
You substantially modify a high-risk system and it remains high-risk – for example through your own retraining that changes performance or purpose.
Art. 25(1)(b)
New intended purpose
You use a system that was not high-risk in a way that makes it high-risk – such as a general assistant for pre-screening job applications.
Art. 25(1)(c)
The third case in particular is underestimated: using a general AI tool for an Annex III purpose can suddenly put the full provider list on you. Check new purposes in advance.
Free template
High-risk checklist as Excel – free, no sign-up
The same list as above, prepared for teamwork: one sheet each for providers and deployers, with columns for owner, status and due date, plus a sheet with all deadlines.
- “Providers” sheet with 17 obligations
- “Deployers” sheet with 14 points
- Status picker and date fields
- “Deadlines” sheet and guide
SimpleAct_High_Risk_AI_Checklist_EN.xlsx
Excel (.xlsx) · status 4 October 2026
Approach
Implementation in seven steps
How to tackle the checklist without getting lost. The order is a recommendation; providers typically need far more lead time for documentation, quality management and conformity assessment than deployers.
- 1
Inventory your AI systems
No overview, no classification. Record all AI systems with purpose, provider and area of use in an AI inventory.
- 2
Classify and clarify your role
Is the system high-risk? Are you provider or deployer – or do you become a provider under Art. 25? Document the result.
- 3
Appoint owners
One owner per system, plus people responsible for oversight, logs, data protection and incidents.
- 4
Record the gaps
Work through the checklist per system and give open points a date. The Excel file helps with distribution.
- 5
Obtain or create documents
Providers create documentation and the declaration; deployers request instructions, declaration of conformity and log details from the provider.
- 6
Enable people
Train overseers, inform staff and workers’ representatives, prepare information for affected persons.
- 7
Monitor operation
Run monitoring, the incident process and log retention – and re-check the list whenever the system changes.
Common mistakes
Where implementation stalls in practice
Role misjudged
Anyone who adapts a system or sells it under their own brand quickly becomes a provider. The role check belongs at the start, not the end.
Waiting for 2027
Technical documentation and quality management are not created in three months. Contracts with providers should secure the documents now.
Oversight on paper only
A named person without training, time and authority does not meet Art. 26(2). Oversight needs competence and backing.
Logs at the provider forgotten
If the logs sit with the provider, access and retention must be agreed by contract – otherwise you can hardly meet your duty.
GDPR treated separately
DPIA, Art. 22 GDPR and information duties run in parallel. Planning them together avoids double work – the FRIA can build on the DPIA.
Changes not followed up
New versions, new data sources, new purposes: every change can affect documentation, risk assessment and even your role.
With SimpleAct
The checklist per AI system – with evidence instead of ticks
In SimpleAct you record your AI systems in the AI inventory, classify them by risk class and work through a compliance checklist per system that matches its risk class. On top, a library of 19 controls along the EU AI Act lets you store status and evidence links per system.
In a pilot project, six AI systems – one of them high-risk – were recorded and assessed in around four hours.
- AI inventory with role, purpose and provider
- Risk classification per system
- Compliance checklist per risk class
- 19 controls with status and evidence links
- DPIA and record of processing in the GDPR module (from Professional)
Open-Source Framework
simpleact-ai-act-checklist
Open-source checklist of all EU AI Act requirements under Art. 8–15 for high-risk AI systems – with article references and evidence guidance.
FAQ
Frequently asked questions about the high-risk AI checklist
More questions? We're happy to help. Send email · Try 7 days (credit card required)
Sources and status
Last reviewed on · SimpleAct editorial team
- Regulation (EU) 2024/1689 (AI Act), EUR-Lex
- Regulation (EU) 2026/1744 (Digital Omnibus), EUR-Lex
- Art. 6 AI Act – Classification as high-risk (German)
- Art. 9 AI Act – Risk management system (German)
- Art. 11 AI Act – Technical documentation (German)
- Art. 13 AI Act – Transparency and instructions for use (German)
- Art. 16 AI Act – Obligations of providers (German)
- Art. 17 AI Act – Quality management system (German)
- Art. 25 AI Act – Responsibilities along the value chain (German)
- Art. 26 AI Act – Obligations of deployers (German)
- Art. 27 AI Act – Fundamental rights impact assessment (German)
- Art. 43 AI Act – Conformity assessment (German)
- Art. 49 AI Act – Registration (German)
- Art. 73 AI Act – Reporting of serious incidents (German)
- Art. 99 AI Act – Penalties (German)
This page, the checklist and the Excel file are a simplified overview and do not constitute legal advice. The text of Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744 is authoritative. The approach, examples and points marked as recommendations are not legal obligations.
Changes to this page
- 4 October 2026 – Rebuilt: interactive checklist by role with task list, free Excel file, evidence and deadline tables, role change under Art. 25, status after the Digital Omnibus. Deployer obligations added; categories now point to the Annex III page.
- 23 May 2026 – First version with obligations from Art. 8 to 18.
Related topics
- Annex III: all high-risk areas with examples
- EU AI Act overview
- EU AI Act risk categories
- All EU AI Act deadlines
- Technical documentation under Annex IV
- AI governance framework
- Deployer obligations under Article 26
- Fundamental rights impact assessment (Art. 27)
- AI in HR and recruiting
- Data protection impact assessment
- Template: AI inventory
- Build an AI register
Keep high-risk obligations in view, system by system
AI inventory, classification, checklist and evidence in one place – for providers and deployers.