Search every question across sub-FAQs

Article 73 requires reporting immediately after the provider establishes a causal link or reasonable likelihood of a link and sets a default outer deadline of 15 days after the provider, or where applicable the deployer, becomes aware of the serious incident. Shorter outer deadlines apply for two categories: a widespread infringement or a serious and irreversible disruption of critical infrastructure management or operation must be reported immediately and no later than two days after awareness, and a death-related incident must be reported immediately after a causal relationship is established or suspected and no later than 10 days after awareness.

If a complete report would delay timely reporting, Article 73 allows an incomplete initial report followed by a complete report. After reporting, the provider must without delay investigate the serious incident and the AI system concerned, including a risk assessment and corrective action, and must cooperate with competent authorities and, where relevant, the notified body.

A deployer is not the normal Article 73 reporter, but it has an explicit escalation duty when it identifies a serious incident: inform the provider first, then the importer or distributor and the relevant market surveillance authorities. Importers and distributors have their own high-risk AI system duties to withhold, notify, or help correct non-conforming or risky systems, so incident intake should route them into the communication record even when the provider owns the Article 73 report.

From 27 July 2026, the AI Office has exclusive supervision over specified AI systems. The covered set includes certain systems based on a general-purpose AI model where the model and system share the same provider or undertaking, subject to the listed exclusions for product-related systems, critical infrastructure, systems provided by law-enforcement or border-management authorities, financial-institution systems within Article 74(6), and justice systems, plus systems that constitute or are integrated into a designated very large online platform or very large online search engine. Providers in that set report Article 73 incidents to the AI Office, which transmits the relevant information to the national market surveillance authority.

Do not merge this system-level route with the EU AI Act rule for providers of general-purpose AI models with systemic risk. Article 55 requires those model providers to keep track of, document, and report without undue delay to the AI Office and, as appropriate, national competent authorities relevant information about serious incidents and possible corrective measures. The Commission's GPAI serious-incident template is for that Article 55 model-provider context, not a replacement for an Article 73 high-risk AI system report.

Article 27 requires the assessment before deployment of a high-risk AI system referred to in Article 6(2), which points to the Annex III high-risk areas. The rule expressly excludes high-risk AI systems intended to be used in the area listed in point 2 of Annex III, the critical-infrastructure area.

The trigger then depends on the deployer. A FRIA is required for deployers that are bodies governed by public law, private entities providing public services, and deployers of high-risk systems in Annex III points 5(b) and 5(c), which cover creditworthiness or credit scoring and risk assessment or pricing for life and health insurance.

The duty applies to the first use. A deployer may rely on a previous FRIA or an existing provider assessment in a similar case, but it remains responsible for checking that the system, process, affected groups, risks, oversight, and mitigations match its own deployment. If a required element changes or becomes outdated during use, the deployer must update the information.

Article 27 gives a concrete assessment list. The record should describe the deployer's process in which the high-risk AI system will be used, the intended period and frequency of use, the categories of natural persons and groups likely to be affected, and the specific risks of harm for those groups.

The FRIA also needs the human oversight implementation described according to the instructions for use, plus the measures to take if the risks materialise. Those measures include internal governance and complaint mechanisms, so the evidence should reach beyond legal sign-off into operating procedures.

A FRIA does not replace a GDPR or law-enforcement data protection impact assessment. From 27 July 2026, amended Article 27(4) lets a deployer cross-reference the relevant parts of a DPIA under GDPR Article 35 or Directive (EU) 2016/680 Article 27, or include those parts in the FRIA, when they already meet an Article 27 obligation. The deployer must still complete every Article 27 element that the DPIA does not cover.

After performing the FRIA, the deployer must notify the market surveillance authority of the results by submitting the filled-out template referred to in Article 27. Separately, Article 49 requires public authorities, Union bodies, agencies, offices, or persons acting on their behalf to register their use of Annex III high-risk systems in the EU database, except Annex III point 2 systems, which are registered nationally.

Article 53 requires providers of general-purpose AI models to keep technical documentation for authorities, provide information to downstream AI system providers, maintain a copyright-compliance policy, and publish a sufficiently detailed summary of the content used for training.

The downstream information must help AI system providers understand the model's capabilities and limitations and comply with their own AI Act obligations, while protecting intellectual property, confidential business information, and trade secrets.

Article 51 classifies a general-purpose AI model as a GPAI model with systemic risk if it has high-impact capabilities, or if the Commission designates it after an ex officio assessment or a qualified alert from the scientific panel using Annex XIII criteria.

The Act creates a compute-based presumption: a GPAI model is presumed to have high-impact capabilities when the cumulative computation used for training is greater than 10^25 floating point operations. Article 52 then requires notification to the Commission without delay and in any event within two weeks after the requirement is met or it becomes known that it will be met. The Commission can also designate a model under the separate Annex XIII route.

Systemic-risk analysis concerns Union-level effects from the model's reach or high-impact capabilities. The Regulation names actual or reasonably foreseeable negative effects on public health and safety, public security, fundamental rights, or society as a whole, including propagation across the value chain. The provider should assess plausible downstream effects and misuse across sectors alongside benchmark performance and direct applications.

Article 55 duties apply in addition to Articles 53 and 54. Providers of GPAI models with systemic risk must evaluate the model with state-of-the-art protocols and tools, conduct and document adversarial testing, assess and mitigate systemic risks at Union level, track and report serious incidents, and protect the model and its physical infrastructure with adequate cybersecurity.

The Article 55 work should be continuous across development, placing on the market, and use. It should cover sources of systemic risk, post-market learning from incidents and misuse, corrective measures, and security risks such as model leakage, unauthorised releases, circumvention of safety measures, unauthorised access, and model theft.

The Article 53 copyright policy and public training-content summary are separate duties. The copyright policy is an internal compliance policy for Union copyright and related-rights law. The public summary is a transparency artifact about the content used for model training, published according to the AI Office template.

Open-source status can remove only the Article 53(1)(a) and (b) documentation and downstream-information duties when the licence and public-availability conditions in Article 53(2) are met. It does not remove the copyright-policy or public training-summary duties, and the exception does not apply to GPAI models with systemic risk.

The Commission guidelines state that Chapter V obligations for GPAI model providers apply from 2 August 2025. They also state that Commission enforcement powers for these obligations enter into application on 2 August 2026, and that providers of GPAI models placed on the market before 2 August 2025 must take necessary compliance steps by 2 August 2027.

The same guidance says the guidelines are not legally binding and that authoritative interpretation of the AI Act is for the Court of Justice of the European Union. Treat the dates above as implementation milestones based on the cited Commission guidance, not as a complete enforcement calendar for every AI Act duty.

Article 72 requires providers of high-risk AI systems to establish and document a post-market monitoring system that is proportionate to the AI technologies and risks of the system. The system must actively and systematically collect, document, and analyse relevant data on the high-risk AI system's performance throughout its lifetime.

The monitoring system must be based on a post-market monitoring plan that forms part of the Annex IV technical documentation. Regulation (EU) 2026/1744, published on 24 July 2026 and entering into force on 27 July 2026, replaces the earlier implementing-act mechanism with Commission guidance, including a template, due by 2 September 2027. Until that guidance is available, an internal or vendor form can organise evidence but should not be presented as the Commission template.

For high-risk systems already covered by Section A of Annex I product legislation, Article 72 allows the provider to integrate equivalent AI Act elements into an existing sectoral post-market system and plan. The same integration option applies to Annex III point 5 systems placed on the market or put into service by financial institutions subject to Union financial-services governance rules.

Deployers are not the Article 72 plan owner, but Article 26 makes them important monitoring inputs. Deployers must monitor high-risk AI systems on the basis of the instructions for use and, where relevant, inform providers under Article 72.

If a deployer has reason to consider that use in accordance with the instructions may present a risk under Article 79(1), the deployer must inform the provider or distributor and the relevant market surveillance authority without undue delay and suspend use. If the deployer identifies a serious incident, it must immediately inform the provider first, then the importer or distributor and the relevant market surveillance authorities; if the provider cannot be reached, Article 73 applies mutatis mutandis.

Provider handling should therefore separate ordinary performance feedback from nonconformity, risk, and serious-incident paths. Article 20 requires providers that consider or have reason to consider their high-risk AI system is not in conformity to immediately take corrective action to bring it into conformity, withdraw it, disable it, or recall it as appropriate.

The evidence should show systematic continuing-compliance review as well as incident logging. Keep the plan, monitored signals, analysis records, outcomes, and technical-documentation updates together so a reviewer can trace why a risk signal did or did not trigger corrective action, a conformity update, or a serious-incident report.

Annex IV also expects technical documentation to describe relevant lifecycle changes and the post-market performance-evaluation system, including the Article 72 monitoring plan. That makes change history part of the post-market monitoring evidence set.

Reopen the Article 72 review when the real-world system no longer matches the monitoring plan, instructions for use, technical documentation, or risk-management assumptions. The strongest triggers are changes that affect intended purpose, high-risk classification, performance, affected groups, human oversight, logs, integration context, or risk controls.

A distributor, importer, deployer, or other third party can become the provider for a high-risk AI system if it puts its name or trademark on the system, makes a substantial modification while the system remains high-risk, or modifies the intended purpose of a non-high-risk AI system so it becomes high-risk. Those lifecycle changes should not be treated as ordinary backlog items.

A provider is the actor that develops an AI system or general-purpose AI model, or has one developed, and places it on the market or puts the AI system into service under its own name or trademark. That test is about development control, commissioning, market placement, service launch, and branding, not only technical authorship.

A deployer is the actor using an AI system under its authority, except for personal non-professional use. A company can therefore be a deployer of a supplier system even when the supplier remains the provider, because the company controls the concrete use context, users, operating process, input data under its control, and internal oversight.