Is AI employee monitoring high-risk under the EU AI Act?

Usually, yes, where the AI evaluates the people it watches. Annex III, point 4(b) lists AI systems intended to be used to monitor and evaluate the performance and behaviour of persons in work-related relationships, and under Article 6(2) that makes them high-risk. The same point also covers AI used to decide on terms of work, promotion or termination, and to allocate tasks based on individual behaviour or personal traits. The Article 6(3) derogation rarely helps, because a tool that assesses an individual worker's performance is in our reading usually profiling, and an Annex III system that profiles natural persons is always high-risk. AI that infers workers' emotions from biometric data such as faces, voices or typing is a different matter: it is not high-risk but, outside medical and safety uses, prohibited under Article 5(1)(f). The answer turns on three things: whether the tool evaluates individuals or only reports aggregates, what it reads (biometric signals bring in Article 5 and Annex III, point 1), and what the output is used for.

This is an information service to help you orient, not legal advice. To see the obligations that attach to your own AI use, answer three questions about it.

Your situation at a glance

Your situation	High-risk?	Why (article / annex point)
AI scores, ranks or flags individual workers on their performance or behaviour	Yes	Annex III, point 4(b), read with Article 6(2). Article 6(3) never applies where the system profiles natural persons
AI recommends or makes decisions on promotion, termination or terms of work	Yes	Annex III, point 4(b): "decisions affecting terms of work-related relationships, the promotion or termination of work-related contractual relationships"
AI allocates tasks or shifts based on workers' behaviour, traits or characteristics	Yes	Annex III, point 4(b): "to allocate tasks based on individual behaviour or personal traits or characteristics"
Scheduling that allocates work on availability or location only	Probably not under 4(b)	Our reading: the task-allocation limb names allocation "based on individual behaviour or personal traits or characteristics"
Productivity dashboard or keystroke/activity tracking that only reports aggregate team figures and evaluates no individual	Probably not, in our reading	Point 4(b) covers systems that "monitor and evaluate" persons. The same tool scoring individuals is the clearer case. Rule-based counting may not be an AI system (Article 3(1))
AI infers workers' emotions, such as anger, anxiety or boredom, from faces, voices or keystrokes	Prohibited, not high-risk	Article 5(1)(f), except where intended for medical or safety reasons. Detecting physical states such as fatigue is not emotion recognition (Recital 18 of the original 2024 text)
Biometric categorisation of workers to infer race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation	Prohibited	Article 5(1)(g)
Other biometric categorisation by sensitive or protected attributes, or emotion recognition that falls within the medical or safety exception	Yes, unless Article 6(3) applies	Annex III, point 1(b) or 1(c)
Face or fingerprint check at the door or clock-in, only to confirm a person is who they claim to be	Not under point 1(a)	Annex III, point 1(a) excludes biometric verification whose sole purpose is that confirmation
CCTV analytics that identify individuals at a distance against a reference database	Yes, unless Article 6(3) applies	Annex III, point 1(a), remote biometric identification (Article 3(41)), in so far as its use is permitted under Union or national law
CCTV analytics that detect intrusions or objects without biometric data or evaluating staff	Not listed, in our reading	No Annex III point covers security analytics as such. If the same footage is used to evaluate workers' behaviour, point 4(b) comes back in
AI that screens, filters or ranks job applicants	Yes, by a different point	Annex III, point 4(a). See the EU AI Act for HR and recruitment software
AI that rates or deactivates platform workers or other self-employed people	Probably yes, in our reading	Point 4(b) covers "persons in such relationships"; Recital 57 names persons providing services through platforms

The legal test

Article 6(2) says that AI systems referred to in Annex III "shall be considered to be high-risk". Annex III, point 4 is headed "Employment, workers' management and access to self-employment". Point 4(b) lists:

"AI systems intended to be used to make decisions affecting terms of work-related relationships, the promotion or termination of work-related contractual relationships, to allocate tasks based on individual behaviour or personal traits or characteristics or to monitor and evaluate the performance and behaviour of persons in such relationships."

That is four use cases in one point. For monitoring, the elements that do the work are:

• "Intended to be used" points to the provider's purpose, as set out in the instructions for use, promotional or sales materials and technical documentation (Article 3(12)). In our reading, a general collaboration suite is not caught merely because an employer could mine it; a tool sold to score worker performance is. An employer that builds its own monitoring model, or turns a general tool to point 4(b) use, needs to check whether it has become the provider (Article 3(3) and Article 25).
• "Monitor and evaluate" is joined by "and". In our reading, a tool that only collects or displays activity data without assessing anyone is a weaker fit than one that turns that data into a judgement about a person. The Act does not draw that line, and a tool sold as monitoring rarely stops at displaying.
• "Performance and behaviour" is broad. Output, speed, attendance, conduct and compliance with procedures are all, in our reading, performance or behaviour.
• "Persons in such relationships" refers back to "work-related relationships". It is not limited to employees (see below).

Point 4(b), unlike points 1, 6 and 7, carries no "in so far as their use is permitted" qualifier, and names no particular user. Recital 57 of the original 2024 text gives the reason for the listing: these systems "may have an appreciable impact on future career prospects, livelihoods of those persons and workers' rights", and "AI systems used to monitor the performance and behaviour of such persons may also undermine their fundamental rights to data protection and privacy."

The Digital Omnibus did not amend Annex III, point 4, or Article 26(7).

Productivity and activity tracking

Keystroke logging, screenshots, application-use tracking and "active time" figures are where most monitoring questions start. Two separate questions arise.

Is it an AI system at all? Article 3(1) defines an AI system as a machine-based system that, among other things, "infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions". A tool that only counts keystrokes and adds them up may not meet that definition. One that uses a model to classify activity as productive or idle, or to predict who is disengaged, is more likely to. See is your software an AI system?.

Does it evaluate individuals? In our reading, an aggregate dashboard that shows a department's average activity, with no individual view and no individual consequence, sits outside the core of point 4(b). A per-person productivity score, a ranking, or a flag that sends a named worker to a manager for review is an evaluation of that person's performance or behaviour. That is the clearer case, and it is also where profiling comes in.

Emotions and biometrics: a different chapter

If a monitoring tool reads faces, voices, posture or the way a person types and draws conclusions about their emotions, it leaves the high-risk analysis and enters the prohibitions. Article 5(1)(f) prohibits:

"the placing on the market, the putting into service for this specific purpose, or the use of AI systems to infer emotions of a natural person in the areas of workplace and education institutions, except where the use of the AI system is intended to be put in place or into the market for medical or safety reasons"

The Commission's guidelines on prohibited practices (C(2025) 5052 final) limit this to inferences drawn from biometric data (paragraph 250), and list keystrokes among behavioural biometrics (paragraph 251). Following Recital 18 of the original 2024 text, they treat physical states such as pain or fatigue as not emotions, so a system that only detects a driver's fatigue to prevent accidents is not emotion recognition (paragraphs 249 and 261). They read "workplace" broadly, "independent from the status as an employee, contractor, trainee, volunteer, etc.", and expressly include self-employment and candidates (paragraph 254). The guidelines are not binding. We cover the line, the narrow medical-or-safety exception and the Commission's worked examples in what Article 5(1)(f) prohibits at work, and do not repeat them here. Where emotion recognition is not prohibited, it falls under Annex III, point 1(c), and is high-risk unless the Article 6(3) derogation applies (the guidelines, paragraph 263, point to the Section 2 requirements), and Article 50(3) requires deployers to inform the people exposed to it.

Article 5(1)(g) separately prohibits "biometric categorisation systems that categorise individually natural persons based on their biometric data to deduce or infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation". A tool that guessed which workers belong to a union from their images would fall here. Other biometric categorisation "according to sensitive or protected attributes or characteristics" is listed as high-risk in Annex III, point 1(b). All of point 1 applies "in so far as their use is permitted under relevant Union or national law".

Security cameras and access control

Annex III has no entry for security analytics as such. In our reading, what matters is what the system does:

• Identifying people remotely. A system that recognises individuals at a distance, without their active involvement, by comparing their biometric data with a reference database is a remote biometric identification system (Article 3(41)) and high-risk under Annex III, point 1(a).
• Confirming identity at a door or terminal. Point 1(a) "shall not include AI systems intended to be used for biometric verification the sole purpose of which is to confirm that a specific natural person is the person he or she claims to be". A face or fingerprint clock-in that only checks the person is who they claim to be is outside point 1(a).
• Detecting events, not people. Intrusion, object or fire detection that processes no biometric data and evaluates no one is not an Annex III use case, in our reading.
• Turning the footage on staff. Once security video is used to assess how workers behave or perform, the purpose has shifted to point 4(b).

The Commission's guidelines note that systems used to recognise aggressive behaviour, for example at a stadium or a station, will often also screen security staff; because their primary objective is not to assess employees' emotions, the guidelines treat them as outside the Article 5(1)(f) prohibition, while deployers must employ safeguards to avoid screening employees and "remain responsible to ensure that employees are not adversely affected by their use" (paragraph 270).

Platform workers and the self-employed

Point 4(b) speaks of "work-related relationships" and "persons in such relationships", not of employees, and the heading of point 4 includes "access to self-employment". Recital 57 says that "relevant work-related contractual relationships should, in a meaningful manner, involve employees and persons providing services through platforms as referred to in the Commission Work Programme 2021". Our reading is that AI used by a platform to rate, rank, allocate work to or deactivate the people providing services through it is within point 4(b). Whether every freelance or business-to-business contract is a "work-related relationship" the Act does not settle.

Note that the Article 26(7) duty to inform (below) is placed on deployers "who are employers" and speaks of "workers". Whether it reaches a platform that is not the employer of the people it monitors is, in our reading, unsettled. The general Article 26(11) duty to inform affected persons does not depend on employment.

The Article 6(3) derogation, and why profiling closes it

Article 6(3) lets an Annex III system escape the high-risk label where it "does not pose a significant risk of harm to the health, safety or fundamental rights of natural persons, including by not materially influencing the outcome of decision making", and it meets at least one of four conditions: it is intended to (a) perform a narrow procedural task; (b) improve the result of a previously completed human activity; (c) detect decision-making patterns or deviations from prior patterns without being meant to replace or influence the previously completed human assessment without proper human review; or (d) perform a preparatory task to an assessment relevant for the Annex III use cases.

The third subparagraph is absolute: an Annex III system "shall always be considered to be high-risk where the AI system performs profiling of natural persons". Article 3(52) borrows the GDPR definition (Article 4(4) GDPR): automated processing of personal data to evaluate personal aspects of a person, "in particular to analyse or predict aspects concerning that natural person's performance at work, economic situation, health, personal preferences, interests, reliability, behaviour, location or movements". "Performance at work" is named first. A system that monitors and evaluates an individual worker's performance or behaviour will usually be profiling, and for that system the derogation is closed.

It stays open, in our reading, for genuinely narrow tools inside point 4: one that transcribes timesheets, checks a form for missing fields or routes a case to the right manager, without evaluating anyone. A provider that relies on it must document the assessment before placing the system on the market or putting it into service (Article 6(4)) and register the system in the EU database (Article 49(2)). See when an Annex III system is not high-risk.

If it is high-risk: who does what, and when

When. The Digital Omnibus amended Article 113. Chapter III, Sections 1, 2 and 3 apply to Annex III systems from 2 December 2027. Under Article 111(2), as amended, and without prejudice to Article 5, the Act applies to operators of high-risk systems placed on the market or put into service before that date only if, from that date, those systems are subject to significant changes in their designs; providers and deployers of high-risk systems intended to be used by public authorities must in any case comply by 2 August 2030. See what the Digital Omnibus changed.

Providers (whoever develops the system and places it on the market or puts it into service under its own name or trademark, Article 3(3)) carry the main load: the Section 2 requirements, a quality management system, a conformity assessment based on internal control under Annex VI, which Article 43(2) provides for points 2 to 8 of Annex III without a notified body, and registration in the EU database (Article 49(1)).

Deployers (the employer using the system under its authority, Article 3(4)) must take appropriate technical and organisational measures to use it in accordance with the instructions for use (Article 26(1)), assign human oversight to people with the necessary competence, training and authority (Article 26(2)), monitor its operation (Article 26(5)) and keep the logs under their control for at least six months unless other law provides otherwise (Article 26(6)). Two duties matter most at work:

• Telling workers first (Article 26(7)). "Before putting into service or using a high-risk AI system at the workplace, deployers who are employers shall inform workers' representatives and the affected workers that they will be subject to the use of the high-risk AI system." The information is to be provided, where applicable, under the Union and national rules and practice on informing workers and their representatives. Recital 92 of the original text says the Regulation is without prejudice to existing obligations to inform or to inform and consult workers.
• Telling the people decided about (Article 26(11)). Deployers of Annex III high-risk systems that make decisions or assist in making decisions related to natural persons "shall inform the natural persons that they are subject to the use of the high-risk AI system".

Right to an explanation. Under Article 86(1), a person subject to a decision taken by the deployer on the basis of an Annex III system's output (point 2 excepted), which produces legal effects or similarly significantly affects them in a way they consider adverse to their health, safety or fundamental rights, has the right to obtain from the deployer "clear and meaningful explanations of the role of the AI system in the decision-making procedure and the main elements of the decision taken". A dismissal or a refused promotion based on a monitoring score is the obvious case. The right yields to exceptions or restrictions under Union or national law (Article 86(2)) and applies only where it is not otherwise provided for under Union law (Article 86(3)). See deployer obligations under Article 26.

Fundamental rights impact assessment: not for most private employers. Article 27(1) requires the assessment from three groups of deployers of Article 6(2) systems: "deployers that are bodies governed by public law, or are private entities providing public services, and deployers of high-risk AI systems referred to in points 5 (b) and (c) of Annex III". Point 4 is not in that list. A private employer using a point 4(b) system is not required to perform one by Article 27, unless it is itself a private entity providing public services. A public-sector employer that is a body governed by public law is.

The national layer: more protection for workers is allowed

Article 2(11) is short and important for employers:

"This Regulation does not preclude the Union or Member States from maintaining or introducing laws, regulations or administrative provisions which are more favourable to workers in terms of protecting their rights in respect of the use of AI systems by employers, or from encouraging or allowing the application of collective agreements which are more favourable to workers."

So the AI Act is a floor for workplace AI, not a ceiling. National employment law, works-council rights and collective agreements may restrict or condition monitoring further. The GDPR also applies in parallel to any monitoring that processes workers' personal data; the AI Act does not replace it.

What applies regardless

• AI literacy (Article 4). Applies since 2 February 2025. As replaced by the Digital Omnibus, it requires providers and deployers to "take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf", such as the managers who read monitoring outputs, and it "does not require providers or deployers to guarantee any specific level of AI literacy of any individual". See the AI literacy duty.
• Prohibited practices (Article 5). Applies since 2 February 2025, including the bans on workplace emotion recognition (Article 5(1)(f)) and sensitive biometric categorisation (Article 5(1)(g)). The two prohibitions the Omnibus added, Article 5(1)(ba) and (bb), apply from 2 December 2026. Breaches can carry fines of up to EUR 35 million or, for an undertaking, up to 7% of its total worldwide annual turnover for the preceding financial year, whichever is higher (Article 99(3)); for SMEs, including start-ups, whichever is lower (Article 99(6)). See prohibited AI practices.
• Transparency (Article 50). Applies since 2 August 2026. Article 50(3) requires deployers of an emotion recognition or biometric categorisation system to inform the people exposed to it; Article 50(1) requires providers to design AI systems that interact directly with people, such as an HR chatbot, so that those people are informed they are interacting with an AI system, unless that is obvious. See transparency obligations.

What to do now

1. List every tool that watches work: activity and keystroke tracking, productivity scoring, call and email analytics, CCTV analytics, access control, task-allocation and scheduling engines, and the monitoring features inside larger HR or collaboration suites.
2. For each, ask whether it evaluates individuals. A score, ranking or flag about a named person points to Annex III, point 4(b). Do not lean on Article 6(3) for anything that assesses a person's performance or behaviour.
3. Check what it reads. Faces, voices, posture or typing patterns used to infer emotions are an Article 5 question now, not a 2027 question.
4. Trace where the output goes. If it feeds decisions on pay, promotion, discipline, dismissal or who gets which work, treat it as in scope.
5. Plan the Article 26(7) information step with your workers' representatives well before 2 December 2027, and check national law and any collective agreement under Article 2(11).

For the general test behind this page, see is your AI system high-risk?. To get a weekly email of verified changes to these rules, join the waitlist.

The official text is Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744. This article is an information service to help you orient — it is not legal advice.