How to Report AI Incidents and Non-Compliance in Germany
If a high-risk AI system causes a serious incident in Germany, the EU AI Act can require the provider to notify the relevant market surveillance authority under Article 73.
The basic deadline is 15 days after the provider or, where applicable, the deployer becomes aware of the serious incident. But the deadline can be much shorter. A widespread infringement or a serious incident involving a serious and irreversible disruption of critical infrastructure must be reported immediately and no later than two days after becoming aware of it. Where a person dies, the report must be made immediately once the provider or deployer establishes or suspects a causal relationship, and no later than 10 days after becoming aware of the incident.
The report goes to the market surveillance authority in the Member State where the incident occurred. In Germany, the Bundesnetzagentur (BNetzA) is the default market surveillance authority under KI-MIG, unless another authority has responsibility under the German allocation of powers. Financial-sector AI can fall under BaFin, while AI systems connected with products covered by Annex I of the AI Act remain linked to the authorities responsible for the relevant product legislation.
This article explains what qualifies as a serious incident, which deadline applies, how reporting works in Germany and what providers need to do after submitting a report.
For the broader German authority structure, see our guide to Germany’s EU AI Act competent authorities.
Need to track AI incident obligations across your systems?
AnnexOps helps AI teams organise obligations, documentation and evidence connected to their AI systems.
What Is a Serious Incident Under the EU AI Act?
Article 3(49) defines a serious incident as an incident or malfunctioning of an AI system that directly or indirectly leads to one of four outcomes:
- The death of a person or serious harm to a person’s health
- A serious and irreversible disruption of the management or operation of critical infrastructure
- An infringement of obligations under Union law intended to protect fundamental rights
- Serious harm to property or the environment
This definition is important because not every technical failure, customer complaint or model error is automatically a reportable serious incident under Article 73.
The connection between the AI system and the outcome can also be direct or indirect. Providers therefore need an incident assessment process capable of identifying potential regulatory significance rather than waiting until the cause is established with absolute certainty.
Article 73 requires reporting once the provider has established a causal link or the reasonable likelihood of such a link.
Who Has to Report a Serious AI Incident?
Article 73 places the reporting obligation on providers of high-risk AI systems placed on the Union market.
The provision also recognises situations where a deployer becomes aware of the incident. The reporting clock can therefore run from the point at which the provider or, where applicable, deployer becomes aware of the serious incident.
The obligation should be built into the provider’s post-market monitoring and incident-management processes. Article 8 of the AI Act specifically requires high-risk AI systems to have procedures related to serious-incident reporting under Article 73.
Reporting Deadlines Under Article 73
The deadline depends on what happened.
| Situation | Article 73 deadline |
| Standard serious incident | Immediately after establishing a causal link or reasonable likelihood of one, and no later than 15 days after becoming aware |
| Widespread infringement | Immediately, and no later than 2 days after becoming aware |
| Serious incident involving critical infrastructure under Article 3(49)(b) | Immediately, and no later than 2 days after becoming aware |
| Death of a person | Immediately after establishing or suspecting a causal relationship, and no later than 10 days after becoming aware |
The two-day deadline is therefore not a general deadline for every severe AI incident. It applies to a widespread infringement and to the specific serious-incident category in Article 3(49)(b), concerning serious and irreversible disruption of critical infrastructure.
For a death, Article 73(4) similarly combines an immediate-reporting requirement with a maximum 10-day outer limit.
Can You Submit an Incomplete Report?
Yes.
Article 73(5) allows the provider or, where applicable, deployer to submit an initial incomplete report when this is necessary to ensure timely reporting. A complete report can follow afterwards.
This is particularly relevant where the applicable deadline is two days or ten days, and the investigation is still developing.
The reporting process should therefore distinguish between:
- Information that is necessary to make the initial notification
- Evidence that must be collected during the investigation
- Additional information required for the final report
Which German Authority Receives the Report?
Article 73 requires providers to report the incident to the market surveillance authority of the Member State where the incident occurred.
Germany’s KI-MIG determines which German authority performs that market surveillance function.
BNetzA: Default Market Surveillance Authority
Under §2(1) KI-MIG, the Bundesnetzagentur is Germany’s competent market surveillance authority unless KI-MIG provides otherwise.
For a high-risk AI system that does not fall within one of the specific exceptions in KI-MIG, BNetzA is therefore the relevant starting point.
The BNetzA’s own AI Act information confirms that providers of high-risk AI systems must report serious incidents to the relevant market surveillance authorities under Article 73.
BaFin: Certain Financial-Sector AI Systems
KI-MIG §2(3) assigns market surveillance responsibility to BaFin for AI systems directly connected with regulated financial activities and placed on the market, put into service or used by specified BaFin-supervised entities.
This means a financial organisation should not automatically assume that every AI Act incident belongs with BNetzA.
The exact allocation depends on the financial activity, the supervised entity and the AI system involved.
Product-Related AI Systems
For high-risk AI systems connected with products covered by the EU AI Act’s Annex I Section A harmonisation legislation, KI-MIG §2(2) assigns market surveillance to the German authorities already responsible for the relevant product legislation.
This matters for areas such as regulated medical devices and other products covered by Annex I.
It is therefore too broad to state that BfArM is automatically the Article 73 reporting authority for all medical-device AI.
For AI systems that are safety components of devices, or are themselves devices covered by Regulations (EU) 2017/745 or 2017/746, Article 73(10) provides a specific reporting route. Serious-incident notifications under the AI Act are limited to the incidents described in Article 3(49)(c), and the report goes to the national competent authority chosen by the Member State where the incident occurred.
This medical-device route also interacts with the reporting systems established under the medical-device legislation. The correct German authority should therefore be confirmed for the specific device and applicable product framework rather than assumed from the AI system alone.
What Information Should the Report Contain?
Article 73 establishes the reporting obligation and deadlines, while the European Commission’s Article 73 materials provide practical guidance and a reporting template for serious incidents involving high-risk AI systems. The Commission published draft guidance and a template for stakeholder consultation in September 2025, and the Commission has since made Article 73 guidance and reporting materials available to support implementation.
The reporting process should be able to identify, at minimum:
- The AI system concerned
- The provider and relevant operators
- The nature and circumstances of the incident
- The applicable serious-incident category
- The date the provider or relevant deployer became aware of the incident
- The established causal relationship or reasonable likelihood of one
- The immediate measures taken
- The investigation and risk assessment
- Corrective actions and follow-up information
The exact information required should be checked against the current Commission reporting template and the instructions of the competent market surveillance authority.
What Happens After You Report?
Submitting the notification does not end the provider’s obligations.
Article 73(6) requires the provider to investigate the serious incident and the AI system without delay. The investigation must include a risk assessment and corrective action. The provider must also cooperate with the competent authorities and, where relevant, the notified body.
There is an important restriction on changing the system during this process.
A provider must not conduct an investigation that involves altering the AI system in a way that could affect a later assessment of the incident’s causes before informing the competent authorities about that action.
This makes evidence preservation an operational requirement, not simply a recommended internal control.
What Does the Market Surveillance Authority Do?
Article 73(8) requires the market surveillance authority to take appropriate measures under Article 19 of Regulation (EU) 2019/1020 within seven days of receiving the notification.
For certain serious incidents involving fundamental-rights infringements under Article 3(49)(c), the market surveillance authority must also inform the national public authorities or bodies referred to in Article 77(1).
National competent authorities must notify the European Commission of serious incidents, whether or not they take action, in accordance with Article 20 of Regulation (EU) 2019/1020.
Special Rules for AI Systems With Other Reporting Obligations
The EU AI Act recognises that some high-risk AI systems already operate under sector-specific reporting regimes.
Article 73(9) limits AI Act serious-incident notifications for certain Annex III high-risk AI systems where providers are already subject to equivalent reporting obligations under Union legislation. In those cases, the AI Act notification is limited to the serious incidents described in Article 3(49)(c).
Article 73(10) contains a specific rule for AI systems that are safety components of devices, or are themselves devices, covered by the EU Medical Devices Regulation or the In Vitro Diagnostic Medical Devices Regulation. The AI Act notification is limited to Article 3(49)(c) incidents and is made to the national competent authority selected by the Member State.
This is why companies operating regulated products should integrate AI incident reporting with their existing sector-specific vigilance and post-market processes rather than create an entirely separate reporting workflow.
Reporting AI Non-Compliance in Germany
Article 73 deals with serious incidents. It should not be confused with the broader mechanisms for reporting suspected non-compliance.
Complaints Under Article 85
Article 85 gives natural and legal persons a right to lodge a complaint with the relevant market surveillance authority where they have grounds to consider that the AI Act has been infringed.
Germany’s KI-MIG establishes BNetzA as the central complaints office. Under §8 KI-MIG, complaints can be submitted to BNetzA, which forwards complaints falling within another market surveillance authority’s jurisdiction to that authority.
A complaint is therefore different from an Article 73 serious-incident notification.
The reporting obligation under Article 73 applies to qualifying serious incidents involving high-risk AI systems. A complaint under Article 85 is a broader mechanism for raising an alleged infringement.
Whistleblower Protection
Article 87 of the EU AI Act connects AI Act infringement reporting with the EU whistleblower protection framework.
Germany’s KI-MIG amended the country’s Hinweisgeberschutzgesetz to include violations of Regulation (EU) 2024/1689 within the relevant whistleblower framework.
This provides a separate reporting route from Article 73. An internal whistleblower report does not replace a provider’s Article 73 notification obligation where a reportable serious incident has occurred.
Practical AI Incident Reporting Checklist
A reliable Article 73 process should be established before an incident occurs.
1. Identify the systems covered
Maintain an inventory of high-risk AI systems subject to Article 73 and record their providers, deployers, intended purposes and applicable regulatory frameworks.
2. Map the responsible authority
For Germany, start with BNetzA but check whether KI-MIG assigns responsibility to BaFin, an existing product-sector market surveillance authority or a Länder authority.
3. Record the awareness date
The reporting deadlines run from when the provider or, where applicable, deployer becomes aware of the serious incident. This date should be recorded as part of the incident record.
4. Classify the incident
Determine whether the event falls within Article 3(49) and whether the two-day, ten-day or standard 15-day deadline applies.
5. Preserve relevant evidence
Maintain logs, system versions, relevant records, incident evidence and investigation material. Avoid changes that could affect the later assessment of causation without first informing the competent authority where Article 73 requires it.
6. Submit the notification on time
If the investigation is incomplete but the reporting deadline is approaching, Article 73(5) allows an initial incomplete report followed by a complete report.
7. Continue the investigation
The provider must investigate the incident, conduct a risk assessment, take corrective action, and cooperate with the competent authorities.
8. Keep the evidence trail
Record the notification, authority communications, investigation findings, corrective measures, and subsequent updates so that the incident can be reconstructed later.
How AnnexOps Helps
Article 73 creates a process that starts with incident detection and continues through reporting, investigation, risk assessment and corrective action.
That process becomes difficult to manage when AI systems, documentation, and evidence are spread across different tools.
AnnexOps helps AI teams maintain a structured AI inventory and connect systems with their applicable obligations, documentation and evidence. For incident management, this can support an internal workflow for recording the affected system, applicable obligation, reporting deadline, authority, investigation records and corrective actions.
The goal is not to replace the competent authority or legal advice. It is to keep the compliance information and evidence associated with the AI system instead of reconstructing the record after an incident occurs.
Need a structured process for AI incident reporting?
AnnexOps helps AI teams organise obligations, deadlines, documentation and evidence across their EU AI Act compliance workflows.
Primary Legal Sources
- Regulation (EU) 2024/1689 — EU AI Act, Article 73: EUR-Lex
- EU AI Act — current consolidated text: EUR-Lex
- KI-MIG — German AI Act implementation and market surveillance law: Gesetze im Internet
- Bundesnetzagentur — High-risk AI systems: BNetzA
- European Commission — Article 73 serious-incident guidance and reporting materials: European Commission
Legal-status note: This article reflects the EU AI Act and German KI-MIG framework available as of September 2026. Article 73 reporting guidance and administrative practice may continue to develop. For an actual incident, organisations should use the current reporting instructions and template of the competent market surveillance authority and obtain appropriate legal advice.
