Skip to product information
1 of 3

Svenska Dokumentmallar

AI Act Article 72 – Post-market monitoring of high-risk AI: plan, signals and corrective actions | Word/PDF/Excel

AI Act Article 72 – Post-market monitoring of high-risk AI: plan, signals and corrective actions | Word/PDF/Excel

File format
DOCX, PDF, XLSX
Document language
5 Swedish work templates Swedish guide English User Guide
Number of files
7

Digital download. No physical product is shipped.

Regular price 79 SEK
Regular price Sale price 79 SEK
Sale Sold out
Taxes included.
View full details

About this template

AI Act Article 72 – post-market monitoring of high-risk AI: plan, signal register and corrective actions. An operational Word/PDF/Excel package for Swedish and EU-based providers of high-risk AI systems that need to document how performance and ongoing compliance are monitored throughout the AI system's lifecycle. Here, the responsible person gets a coherent workflow from legal applicability assessment to actual evidence of follow-up, periodic decision and changed system version.

Important limitation: The package is an independent preparatory working document. It is not the European Commission's official monitoring plan, not an automatic AI monitoring system, not a safety certification and not a regulatory notification. Article 72 applies to providers of covered high-risk AI systems, not automatically to all companies that use AI.

Quick facts: content and scope

Price SEK 79 · 15 files in a digital ZIP. Seven editable Word files (DOCX), seven corresponding searchable and printable A4 PDFs and an Excel workbook with nine tabs. The package contains five Swedish working templates, a Swedish user guide and a separate English User Guide with instructions for the Swedish files. The PDF versions are not interactive forms. No file makes API calls, reads an AI model or performs testing in the customer's system.

01 · Package contents

Five Swedish working templates: (1) applicability and provider responsibility, (2) post-market monitoring plan with data sources, (3) registration of monitoring signal and actual evidence, (4) corrective action and separate incident escalation, and (5) periodic review, model change and version decision. Also a Swedish guide, English user guide and a blank nine-tab Excel workbook. Total 7 DOCX + 7 PDF + 1 XLSX = 15 customer files.

02 · What the files contain

Traceable system IDs, provider role, model and plan version, classification according to Article 6, monitoring objectives, data sources, source and evidence IDs, observations, risk assessment, legal escalation decisions, corrective actions, verification evidence and periodic decisions. Excel shows completion requirements or basis for manual control – never automatically "AI-approved" or "incident reported".

Which companies need post-market monitoring for AI?

The product is aimed at providers that release or put high-risk AI systems into service and need to follow up on risks and performance during the lifecycle. Example: A system provider documents after launch that performance has changed, collects feedback from affected users and needs to assess corrective actions. Another example is a provider that has updated its model and must feed observations back into documentation, risk management and controls.

Do not choose the product if you only want a general AI policy, need a legal system approval, lack a defined high/non-high-risk assessment or need a standalone system to actually collect technical logs from the production environment. It cannot replace the provider's technical and compliance functions.

File list – exactly what is included?

File Main task Delivery
01 · Applicability and responsibility Assess provider/deployer, Article 6, Annex I/III, applicability, transition and integration possibility. Swedish · DOCX/PDF
02 · Post-market plan and data sources Plan ID, objectives, metrics, data sources, analysis frequency, confidentiality and link to Annex IV. Swedish · DOCX/PDF
03 · Signal, investigation and evidence log Detected deviation, source, actual consequence, relevant requirements, triage, risk and decision evidence. Swedish · DOCX/PDF
04 · Corrective action and escalation Action decision, verified outcome and separate assessment of incident reporting according to Article 73. Swedish · DOCX/PDF
05 · Review and version decision Periodic follow-up, changed AI model, open signals, documentation update and attestation. Swedish · DOCX/PDF
06 · Swedish user guide Workflow, legal boundaries, transition issues, data protection and links to EU sources. Swedish · DOCX/PDF
07 · English User Guide English instructions for the Swedish working forms; the forms are not translated. English · DOCX/PDF
08 · Excel – follow-up register Nine tabs with systems, plans, data sources, observations, investigations, actions and decisions. Swedish · XLSX

Excel: nine functional tabs

Overview, System, Plans, Data Sources, Observations, Investigations, Actions, Decisions and Read Me. The overview only counts actually filled-in entries. The customer file is empty upon delivery and lacks test and personal data. Plans, observations, investigations and actions have administrative formulas. Missing data is displayed as COMPLETE BASIS. When predefined data is specified, only BASIS FOR MANUAL CONTROL is displayed. The system does not make automatic legal assessments, perform no ML evaluation and does not connect to external services.

How do you work with the post-market plan?

  1. First assess the correct role and applicability. Identify the provider and the actual high-risk classification according to Article 6; also check any product and sector regulations.
  2. Collect relevant technical documentation. Link the plan to the risk file, product version, user instructions and Annex IV where applicable.
  3. Establish proportionate follow-up. Describe actual data sources, relevant signals, measurement methods, responsibility and periodicity.
  4. Record observations with genuine evidence. Distinguish provider feedback, signal, error, security risk and suspected serious incident.
  5. Assess and document actions. Separate incident assessment according to Article 73 should be done when circumstances require it.
  6. Verify effect and version. Keep apart the decided action, actual execution, test result and remaining risk.
  7. Archive manual conclusion. Update the monitoring plan, risk file and technical documentation with the actual source and responsible person.

Legal boundaries according to EU AI Act

Articles 72.1–2 require a proportionate, documented monitoring system that actively and systematically collects, documents and analyzes relevant performance and compliance information during the lifecycle. Where relevant, interaction with other AI systems should be considered. Sensitive operational data from law enforcement users is not covered by the current obligation under paragraph 2.

Article 72.3 – new wording 2026: The plan shall be part of the technical documentation according to Annex IV. The Commission shall adopt guidance, including a plan template, by September 2, 2027 at the latest according to amending Regulation (EU) 2026/1744. This package does not claim to be the future official template.

Article 72.4: For certain AI systems that are already covered by product-specific Union legislation according to Annex I Section A and for certain financial systems according to Annex III point 5, Article 72 can be integrated into existing follow-up, under the specified conditions of equivalent protection. This does not mean a general exemption from compliance.

Date and role: The EU AI Act has different application dates. Chapter III Section 1–3 applies from December 2, 2027 for Annex III systems and from August 2, 2028 for covered Annex I systems according to the current consolidated legal act. Article 72 is found in Chapter IX; do not automatically conclude that exactly the same deadline applies to every obligation without assessing Articles 111 and 113. Article 26 deals with the user's special obligations, while Article 73 applies to separate reporting requirements for serious incidents.

Difference from other Mallbutiken templates

AI Governance covers overall AI governance. AI provider model change focuses on a specific change event. AI incident and evidence deals with incident investigation and human review. This Article 72 package, however, is intended for the provider's ongoing, systematic follow-up of high-risk AI after launch, with plan, data sources, signals, action follow-up and version evidence in a coherent register.

Frequently asked questions

Does Article 72 apply to every company that uses ChatGPT?

No. The provision applies to providers of high-risk AI systems covered by the legal act. Role and classification must be determined individually.

Is this the official EU plan template?

No. The Commission's guidance including template shall be adopted by September 2, 2027 at the latest according to amended Article 72.3. Mallbutiken's product is a standalone preparatory document.

Can a product provider use an existing medical device or other monitoring plan?

In some cases, yes. Article 72.4 enables integration into already established systems according to specially specified regulatory areas when equivalent protection is ensured.

Must the user submit all logs to the provider?

No. Sources must be relevant and handled legally. Article 72.2 also exempts sensitive operational data from law enforcement users.

Does Excel send incident reports according to Article 73?

No. Excel contains no connection to authorities or automatic deadline assessment. The actual incident type and possible notification are assessed separately.

Are all working documents in English?

No. The five working templates and Excel are in Swedish. A separate English User Guide explains the Swedish documents. The PDFs are searchable printed copies without form fields.

Legal sources, revision and responsibility

Legal sources: AI Regulation (EU) 2024/1689, consolidated July 2026; 2026/1744; Article 72; Articles 16, 26, 73, 111 and 113 as well as Annex IV. Version 1.0 · November 16, 2026. AI-assisted source control, PDF/Word file testing, Excel scenario testing and ZIP integrity check have been performed. No external lawyer, authority or certifier has approved the templates. The authorized responsible person must adapt and review every plan against the actual AI system, changed rules and upcoming EU guidance.