Join our Newsletter — 33% off our NHI Course
Home› Guides› EU AI Act Article 26 Deployer Obligations Guide:…
Guide AI Regulation & Law

EU AI Act Article 26 Deployer Obligations Guide: Logs, Monitoring, Worker Notice and Oversight

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, under the editorial direction of Lalit Choda Updated 18 November 2026 17 min read
On this page

Article 26 of the EU AI Act lists twelve duties that fall on every organisation using a high-risk AI system under its own authority, and most of them land on security, identity and platform teams rather than on legal. This guide works through Article 26 paragraph by paragraph: what each duty requires in the text of Regulation (EU) 2024/1689, what evidence an authority is likely to ask for, and how to build the controls into the way you run AI systems, service accounts and AI agents.

If you still need to decide whether you are a provider or a deployer, or whether your changes have turned you into a provider under Article 25, start with the EU AI Act Provider vs Deployer Obligations Guide. This page assumes you have settled that question and are a deployer of a high-risk system.

This guide is not legal advice. Status as of 17 Nov 2026: it reflects Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744 (the Digital Omnibus on AI). Confirm the position for your systems with qualified counsel.

Key takeaways

  • Article 26 applies to deployers of high-risk AI systems. After the Digital Omnibus on AI, Chapter III Sections 1 to 3, which include Article 26, apply from 2 December 2027 for Annex III systems and from 2 August 2028 for Annex I product systems.
  • The core duties are operational: use the system according to the provider's instructions, assign competent human oversight, control input data quality, monitor operation, keep logs for at least six months, and stop using a system you believe presents a risk.
  • Employers must inform workers' representatives and affected workers before putting a high-risk system into service at the workplace (Article 26(7)).
  • Deployers of Annex III systems that make or assist decisions about people must tell those people that the system is being used (Article 26(11)), and Article 86 gives affected persons a right to an explanation from the deployer.
  • Breaching Article 26 can attract fines of up to EUR 15,000,000 or 3% of worldwide annual turnover, whichever is higher (Article 99(4)), with the lower of the two applying to SMEs.
  • Article 26 evidence is mostly identity and access evidence: who may use, oversee, feed and stop the system, and logs that attribute each action. See the EU AI Act for AI Agents map.

Article 26 at a glance

ParagraphDuty in shortTypical ownerEvidence to hold
26(1)Technical and organisational measures to use the system in line with the instructions for usePlatform or service ownerConfiguration baseline mapped to the instructions for use
26(2)Human oversight by people with competence, training, authority and supportBusiness owner with HR and IAMNamed oversight roles, training records, role assignments
26(3)Other legal duties are unaffected; deployer may organise its own resources for oversightLegal and complianceOversight operating model
26(4)Input data relevant and sufficiently representative, where the deployer controls itData ownerData source register and quality checks
26(5)Monitor operation, inform the provider, suspend on risk, report serious incidentsSecurity operations and service ownerMonitoring runbook, incident records, provider contacts
26(6)Keep automatically generated logs under the deployer's control for at least six monthsSecurity engineeringRetention configuration, access control on logs
26(7)Inform workers' representatives and affected workers before workplace useHR and works council liaisonDated notices and consultation records
26(8)Public authorities: register use under Article 49; do not use unregistered systemsPublic sector deployersEU database registration reference
26(9)Use provider information for a GDPR data protection impact assessment where applicableData protection officerDPIA referencing the instructions for use
26(10)Strict conditions for post-remote biometric identification in criminal investigationsLaw enforcement deployersAuthorisations, police file entries, annual reports
26(11)Inform natural persons subject to Annex III decisionsProduct and customer teamsNotice text and delivery logs
26(12)Cooperate with competent authoritiesComplianceRegulator contact procedure

Scope: who Article 26 applies to and from when

Article 26 binds deployers of high-risk AI systems. A deployer is the natural or legal person, public authority, agency or other body using an AI system under its authority, except for personal non-professional activity. A system is high-risk either because it is a safety component of, or is itself, a product covered by Annex I legislation that needs third-party conformity assessment (Article 6(1)), or because it falls within an Annex III use case such as employment, credit scoring, access to essential services, education or critical infrastructure (Article 6(2)), unless the provider has validly concluded under Article 6(3) that it does not pose a significant risk. The EU AI Act Article 6(3) Guide explains that test.

Timing changed in 2026. Regulation (EU) 2026/1744, adopted on 8 July 2026 and in force from 27 July 2026, amended Article 113 so that Chapter III Sections 1 to 3 apply from 2 December 2027 for systems under Article 6(2) and Annex III and from 2 August 2028 for systems under Article 6(1) and Annex I. Article 26 sits in Section 3, so these are the dates for deployer duties. For high-risk systems intended for use by public authorities, the amended Article 111(2) gives providers and deployers until 2 August 2030 to comply where the transitional rule applies.

Two adjacent duties apply to many deployers sooner or alongside Article 26. The AI literacy measure in Article 4 applies already (see the EU AI Act AI Literacy Obligations (Article 4) Guide), and certain deployers must complete a fundamental rights impact assessment under Article 27 before first use (see the EU AI Act Fundamental Rights Impact Assessment).

Paragraph 1: using the system according to the instructions for use

Article 26(1) requires deployers to take appropriate technical and organisational measures to ensure they use high-risk systems in accordance with the instructions for use that accompany them, subject to paragraphs 3 and 6. The instructions for use are the provider's Article 13 deliverable, so the deployer's first job is to obtain them, version them and read them as an operating specification. The EU AI Act Article 13 Guide lists what they must contain.

In practice, "in accordance with" means three things for a security team:

  • Configuration conformance. Translate every configurable statement in the instructions (thresholds, permitted input types, supported languages, required human review steps, integration patterns) into a configuration baseline, and check deployed settings against it. Drift is a compliance event, not only an operational one.
  • Purpose conformance. Restrict who and what can invoke the system to the intended purpose. A credit-scoring model exposed through a general internal API that any service account can call will soon be used outside its intended purpose. Use scoped client credentials, per-consumer authorisation and an approval step for new integrations.
  • Change control. Using the system outside its intended purpose, or modifying it substantially, can make you a provider under Article 25. Route model updates, prompt template changes, new connectors and new agent tools through change approval that asks the Article 25 question.

Paragraphs 2 and 3: assigning human oversight

Article 26(2) requires deployers to assign human oversight to natural persons who have the necessary competence, training and authority, as well as the necessary support. Article 26(3) adds that paragraphs 1 and 2 do not affect other deployer obligations under Union or national law, and that the deployer is free to organise its own resources and activities to implement the oversight measures the provider indicates.

The provider designs the oversight measures under Article 14; the deployer staffs them. Three tests help show the assignment is real:

  • Named people, not teams. Record the individuals, or the role and current holder, responsible for oversight of each system.
  • Competence and training. Link each overseer to training on the system's capabilities and limitations, automation bias and the stop or override procedure. This overlaps with Article 4 AI literacy measures.
  • Authority. Overseers must be able to override, disregard or stop the system. Make that a technical permission: a role in your identity provider that grants the override or kill switch, assigned through your access request process and reviewed periodically.

The EU AI Act Article 14 Guide covers the design side. For agentic systems, separate the identity that approves a high-impact action from the agent's own identity, so that oversight cannot be bypassed by the agent acting with the approver's token.

Paragraph 4: input data under the deployer's control

To the extent the deployer exercises control over the input data, Article 26(4) requires it to ensure that input data is relevant and sufficiently representative in view of the intended purpose of the system. This is narrower than the provider's Article 10 data governance duty, which covers training, validation and testing data, but it is easy to breach in production.

Controls that work:

  • Keep a register of every input source feeding each high-risk system: databases, APIs, document stores, retrieval indexes and upstream agents, with the service identity that reads each one.
  • Validate inputs at the boundary against the types and ranges the instructions for use describe, and log rejects.
  • Control write access to retrieval sources. If anyone with broad permissions can change the documents a system relies on, you cannot show the input data is relevant or representative, and you have a data poisoning risk.

Paragraph 5: monitoring, suspension and serious incidents

Article 26(5) is the paragraph most often underestimated. Deployers must monitor the operation of the system on the basis of the instructions for use and, where relevant, inform providers in line with Article 72 on post-market monitoring. Where deployers have reason to consider that using the system in accordance with the instructions may result in it presenting a risk within the meaning of Article 79(1), they must inform the provider or distributor and the relevant market surveillance authority without undue delay, and suspend use. Where deployers identify a serious incident, they must immediately inform first the provider, and then the importer or distributor and the relevant market surveillance authorities. If the deployer cannot reach the provider, Article 73 applies with the necessary adjustments. The obligation does not cover sensitive operational data of law enforcement deployers.

Financial institutions subject to internal governance requirements under Union financial services law are deemed to meet the monitoring obligation by complying with those governance rules.

What to build:

  • A monitoring runbook per system that states the signals to watch (error rates, override rates, complaint volumes, drift indicators, unexpected tool calls by agents), thresholds and who reviews them.
  • A provider contact route agreed in the contract, with a named mailbox or portal and a fallback, so "immediately" is achievable at 3am.
  • A suspension mechanism that can be operated quickly: disable the system's service credentials, revoke its API keys, pause agent schedules or route traffic to a manual process. Test it.
  • Integration with incident response. Add an "AI serious incident" classification to your incident process so that the regulatory notification clock is triggered from the same triage that handles security incidents. The EU AI Act Post-Market Monitoring Guide (Article 72) explains what providers do with the information you send.

Paragraph 6: keeping logs for at least six months

Article 26(6) requires deployers to keep the logs automatically generated by the high-risk system, to the extent those logs are under their control, for a period appropriate to the intended purpose of the system and of at least six months, unless applicable Union or national law, in particular on personal data protection, provides otherwise. Financial institutions keep the logs as part of the documentation kept under Union financial services law.

The provider must design the system to produce logs under Article 12 (see the EU AI Act Article 12 Guide). The deployer's question is which of those logs it controls. For a self-hosted system, that is usually all of them. For a SaaS system, it may be only what the provider exposes through an export or API, and the contract should say how long the provider keeps the rest and how you can obtain it.

Practical points:

  • Set retention to the longer of six months and your other obligations, then reconcile with GDPR storage limitation. Record the reasoning.
  • Protect log integrity. Write logs to storage the system's own identity cannot modify or delete, so a compromised or misbehaving agent cannot erase its trail.
  • Record the calling identity (user, service account or agent), input reference, output, any override and the configuration version. Control access to logs. They often contain personal data and prompts. Grant read access by role and log access to the logs.

Paragraph 7: informing workers before workplace use

Before putting into service or using a high-risk AI system at the workplace, deployers who are employers must inform workers' representatives and the affected workers that they will be subject to the use of the system. The information must be provided, where applicable, in accordance with Union and national rules and practice on informing workers and their representatives.

Many Annex III employment use cases trigger this: recruitment screening, task allocation, performance monitoring and decisions on promotion or termination. Points to manage:

  • The duty is "before", so build it into the deployment gate. No production go-live for a workplace high-risk system without a dated notice record.

Paragraphs 8 to 10: public authorities, DPIAs and biometric identification

Registration for public authority deployers

Under Article 26(8), deployers that are public authorities, or Union institutions, bodies, offices or agencies, must comply with the registration obligations in Article 49. If they find that the system they plan to use is not registered in the EU database referred to in Article 71, they must not use it and must inform the provider or distributor. Build a registration check into procurement.

Data protection impact assessments

Article 26(9) requires deployers, where applicable, to use the information provided under Article 13 to carry out a data protection impact assessment under Article 35 of Regulation (EU) 2016/679 or Article 27 of Directive (EU) 2016/680. Treat the instructions for use as a required input to the DPIA, and cross-reference the two documents. The EU AI Act vs GDPR page maps the overlap.

Post-remote biometric identification

Article 26(10) sets strict rules for deployers of high-risk systems for post-remote biometric identification in a criminal investigation: prior authorisation by a judicial or binding administrative authority, or a request no later than 48 hours after use; use limited to what is strictly necessary for a specific criminal offence; immediate stop and deletion if authorisation is refused; no untargeted use and no adverse decision based solely on the output; documentation in the police file and annual reports. Member States may go further.

Paragraphs 11 and 12: informing people and cooperating with authorities

Without prejudice to Article 50, deployers of Annex III high-risk systems that make decisions, or assist in making decisions, related to natural persons must inform those persons that they are subject to the use of the system. For law enforcement purposes, Article 13 of Directive (EU) 2016/680 applies. This is separate from the Article 50 transparency duties covered in the EU AI Act Article 50 Transparency Obligations Guide.

Pair this notice with Article 86. Any affected person subject to a decision based on the output of an Annex III high-risk system (other than point 2 critical infrastructure systems) that produces legal effects or similarly significantly affects them in a way they consider adverse to their health, safety or fundamental rights can obtain from the deployer clear and meaningful explanations of the role of the system in the decision procedure and the main elements of the decision. You can only explain what you logged, which is another reason to keep decision-level records under paragraph 6.

Article 26(12) requires deployers to cooperate with the relevant competent authorities in any action they take in relation to the system to implement the Regulation. Prepare by naming a regulatory contact and keeping an evidence pack per system: instructions for use, configuration baseline, oversight assignments, log retention settings, notices, DPIA and incident records.

Enforcement and penalties for deployers

Market surveillance authorities designated under Article 70 supervise deployers. Under Article 99(4), non-compliance with the obligations of deployers under Article 26 is subject to administrative fines of up to EUR 15,000,000 or, for an undertaking, up to 3% of total worldwide annual turnover for the preceding financial year, whichever is higher. For SMEs, including start-ups, each fine is capped at whichever of the percentage or amount is lower (Article 99(6)), and the Digital Omnibus added an equivalent rule for small mid-cap enterprises (new Article 99(6a)). Supplying incorrect, incomplete or misleading information to authorities attracts up to EUR 7,500,000 or 1% (Article 99(5)). Article 99(7) lists the factors authorities weigh, including cooperation, the technical and organisational measures taken, and whether the deployer notified the infringement itself. See the EU AI Act Penalties and Enforcement Guide for the full regime.

Common failure patterns

  • Treating the provider's CE marking as the end of the story, with no deployer-side configuration baseline or monitoring.
  • Oversight assigned to a shared mailbox or a team name, with no individual holding override authority in the identity system.
  • Logs retained only by the SaaS provider for 30 days, with no contract term or export to meet the six-month minimum.
  • Logs writable or deletable by the AI system's own service account.
  • No tested way to suspend the system quickly, so "suspend use" in paragraph 5 means a change request that takes a week.
  • Integrations added by other teams through shared API keys, taking the system outside its intended purpose and possibly triggering Article 25.

Practitioner checklist

  • Inventory every high-risk AI system you deploy, with the provider, intended purpose, Annex reference and the date Article 26 applies to it.
  • Obtain and version the instructions for use and derive a configuration baseline from them.
  • Restrict invocation to approved identities and integrations, and route new integrations through an Article 25 check.
  • Name the human overseers, record their training and grant override and stop rights through a reviewed role.
  • Register input data sources and the identities that read and write them; validate inputs at the boundary.
  • Write a monitoring runbook, agree provider contact routes and test the suspension mechanism.
  • Add an AI serious incident category to incident response with the notification sequence from Article 26(5).
  • Set log retention to at least six months, protect integrity and restrict access.
  • Issue and record worker notices before any workplace high-risk deployment.
  • Check EU database registration if you are a public authority, and complete or update the DPIA using the Article 13 information.
  • Publish notices to affected persons for Annex III decisions and prepare an Article 86 explanation process.
  • Keep an evidence pack per system for authority requests under Article 26(12).

Standards and references

Related NHI Mgmt Group resources: EU AI Act Provider vs Deployer Obligations Guide · EU AI Act High-Risk AI Systems Guide · EU AI Act Post-Market Monitoring Guide (Article 72) · EU AI Act Compliance Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick both boxes to continue.