Essential Services

Automated decisions determining eligibility

• An AI system intended to decide on the grant of unemployment benefits in individual cases. The system analyses whether the person qualifies for the benefit, during which period and in which amount, and makes a recommendation to a case handler about the eligibility and the receiving of the benefits. Such an AI system is high-risk, because it is intended to evaluate a natural person’s eligibility, and makes a recommendation on the granting or denial of these benefits.
• An AI system intended to assess a natural person’s eligibility for social benefits following an application. The system processes application data and cross-references external databases, after which it generates a decision on the applicant’s eligibility for social benefits based on the given data. Such an AI system is high-risk, because it is intended to evaluate a natural person’s eligibility and generates a decision on their elgibility.
• An AI system intended to be used by public administrations to evaluate a natural person’s eligibility for social housing or unemployment benefits through a scoring system that is not a prohibited practice covered by Article 5(1)(c) AI Act. The system draws on demographic, behavioural, and financial data to predict the likelihood of fraud or long-term dependency. Such an AI system is high-risk, since it aids in the evaluation of a natural person’s eligibility and materially influences decision about whether the applicant will receive the benefits and services.

Prioritisation

• An AI system intended to evaluate applications for social housing. The system evaluates the profiles of applicants for eligibility, urgency level and prioritisation for specific social housing units based on income, household size, age, employment status and location. The system can substitute or influence human decision-making and is high-risk, since it has a direct impact on access to essential housing benefits.
• An AI system intended to prioritise the allocation of home-care services. Such systems may be used by local authorities to determine the distribution and prioritization of limited in-home care hours among eligible persons. Such an AI system is high-risk, since it materially influences the access to essential care services for elderly persons or persons dependent on such care, thereby potentially affecting their fundamental rights and well-being.
• An AI system intended to assess the eligibility of applications for access to unemployment support or income support schemes, administered by public authorities, and prioritise those applications accordingly in a context of limited resources or processing capacity. Such an AI system is high-risk, since it evaluates applications for essential public assistance benefits and services and it materially influences the granting, timing or effective access to such benefits and services through prioritisation.

Supportive tools for human review determining eligibility or access

• An AI system intended to analyse applications for essential public assistance benefits submitted by natural persons (e.g. sickness benefits) in order to detect potential irregularities and/or inaccuracies, or fraud indicators and to flag applications for further investigation by a public authority. When an application is flagged, a human case handler takes over to assess whether the claim is fraudulent. Such an AI system is high-risk, since it is used in the context of evaluating eligibility for essential public assistance benefits and materially influences access to such benefits. In particular, flagging determines whether applications are subject to further scrutiny, delay, suspension or potential refusal and is therefore instrumental to decisions affecting the granting, continuation or effective access to those benefits.
• A chatbot intended to answer legal questions of a case handler that are specifically related to the evaluation of an application of a natural person to receive essential care benefits or services administered by a public authority. The case handler can grant or refuse the benefits to which the natural person applied, while remaining responsible for the final decision, based on the answers given by the chatbot that are personalised to the specific case/individual eligibility assessment. Such an AI system is high-risk, since the case handler’s decision on the eligibility of the natural person to receive care benefits or services is materially influenced by the legal answers given by the chatbot.
• An AI system is intended to assess applications for healthcare in order to determine whether an in-depth examination of the application should be carried out by an external body. Even if the decision to forward is taken by a case handler, such an AI system is high-risk, since the system preliminary assessment acts as a decisive step for further examination determining the access to healthcare services.

Proactive invitation with an eligibility check

• An AI system intended to be used by public authorities for health preventive screenings to suggest personalised invitations for preventive screening (such as mammograms), based on population segmentation, performing also an evaluation of the eligibility for the preventive screening. Such an AI system is high-risk, since it evaluates the eligibility of a person to do a preventive screening (a healthcare service), even if this is done through a proactive invitation.

Proactive invitations without an evaluation of the eligibility

• AI systems intended to proactively identify persons in need of preventive care. Such an AI system falls outside the use case of point 5(a) of Annex III, since it only identifies persons in need of preventive care and does not determine the eligibility of the identified persons, provided that the system’s outputs are not used to prioritise, exclude or otherwise influence access to such a service.
• AI systems intended to match persons with the public support services that fit them best based on their needs, using anonymised data to suggest and pre-fill applications for support services. Such an AI system falls outside the use case of point 5(a) of Annex III, since it is not intended to evaluate the eligibility of a natural person, but to support the delivery of public support services.

Case-handler allocation

• AI systems intended to merely inform by which body or competent person an application for essential public assistance benefits is (further) to be assessed. Such an AI system falls outside the use case of point 5(a) of Annex III, since it only recommends who should assess the application, without an actual evaluation of the eligibility.

Eligibility checks and reimbursement of costs of legal persons 

• AI systems intended to assess applications or claims from companies for reimbursement of costs, but not from natural persons. Such an AI system falls outside the use case of point 5(a) of Annex III, since it does not assess applications or claims from natural persons.

Chatbots answering factual questions of a case handler 

• A chatbot intended to answer factual questions of a case handler that are related to the evaluation of an application of a natural person to receive healthcare benefits (such as the age of the natural person). The case handler can grant or deny the benefits to which the natural person applied, based on the answers given by the chatbot. Such an AI system falls within the use case of point 5(a) of Annex III, since the case handler’s decision on the eligibility of the natural person to receive healthcare benefits depends on the answers given by the chatbot. However, since the system only provides existing and factual information in a structured manner, without evaluating this information for the eligibility of a natural person’s application for healthcare benefits, and the case handler must review the information made available by the chatbot to ensure that the decision is not based solely on the output of the AI system, it falls under the exception for AI systems intended to perform narrow procedural tasks listed in Article 6(3)(a) and (d) AI Act.

Supportive tools for human review

• AI systems intended to summarise medical reports which will be used by a case handler as the basis for their decision to grant or deny healthcare benefits. Such an AI system falls within the use case of point 5(a) of Annex III, since it is intended to be used for the access to healthcare services. However, since the system is limited to summarising the information in the reports and the case handler must review the relevant information in the reports before taking a decision, it falls under the exception for AI systems intended to perform narrow procedural tasks listed in Article 6(3)(a) and 6(3)(d) AI Act.

Language tools

• AI systems intended to translate applications for public assistance services made in a foreign language. Such an AI system falls within the use case of point 5(a) of Annex III, since it is intended to be used for the evaluation of a natural person’s application for public assistance benefits. However, since the translation of applications for assistance benefits and services is an unavoidable step in the process, but it is not decisive for the evaluation of the eligibility of those applications, the system falls under the exceptions for AI systems intended to perform narrow procedural and preparatory tasks listed in Article 6(3)(a) and (d) AI Act.
• AI systems intended to be used in conversations between a case handler and a natural person who wants to apply for public assistance services and benefits by converting speech into text. The system creates a summary of the conversation that will be used to make a decision concerning the eligibility of a natural person. Such an AI system falls within the use case of point 5(a) of Annex III, since it is linked to the evaluation of a natural person’s application for public assistance benefits and services. However, in so far as the case handler is present during the conversation and is able to check the summary made by the system, so that the output is only a supporting but not decisive element for his or her decision, the system falls under the exception for AI systems intended to perform only narrow procedural tasks listed in Article 6(3)(a) and 6(3) (d) AI Act.

Credit scoring for consumer lending and mortgages

• An AI system intended to create a numerical representation of a natural person's creditworthiness, based on their payment behaviour and income, that is intended to support the decision-making on a consumer credit or a mortgage, which qualify as essential private services.

Credit scores used by third parties other than the deployer 

• An AI system intended to establish a credit score deployed by a credit agency, whose resulting scores are shared with a third party for decisions whether to grant a loan/mortgage or access to housing, healthcare or telecommunication services to a natural person. To fall within the use case of point 5(b) of Annex III, it is not necessary that the deployer of the system is identical to the party using the credit score for its decision-making.

Customer classification and personalised marketing

• An AI system intended to classify customers, for example, to fulfil information obligations, to provide tailored information to customers, to assess the suitability of a product or to make personalised marketing offers, so long as the classification does not play a part in the assessment of the creditworthiness of a natural person.
• An AI system intended to perform advanced segmentation to gain a deeper understanding of customer groups and how they use certain services based on demographic and behavioural data. The same may be true for pricing simulations, that are used to test how changes in pricing might affect customer behaviour, competitiveness or profitability, if they are distinct from the AI system carrying out a creditworthiness assessment. Even though their outputs may be used further downstream, they represent an earlier step in the value chain.

Support before or after a credit decision

• An AI systems intended for customer support related to the assessment of their creditworthiness should not be classified as high-risk under point 5(b) of Annex III, since they do not assess the creditworthiness or establish a credit score of a natural person. These AI systems may assist applicants in understanding or completing the credit application form (e.g. by explaining terminology) or provide dynamic feedback on how specific answers may influence the likelihood of approval. If they are not intended to be used as part of the creditworthiness assessment or credit-scoring process, since they merely support the applicants to prepare their credit application but do not participate in the formal credit-scoring assessment, they fall outside the use case of point 5(b).
• An AI system intended to be used to handle and manage complaints following a decision on loans or health and life insurance, so long as it does not at the same time constitute an evaluation of the creditworthiness or the establishment of a credit score. Such systems usually represent a separate stage of the credit application process where individuals challenge a decision on the credit provision or creditworthiness assessment.

Monitoring of credit exposure

• An AI system intended solely for monitoring credit exposures for internal prudential purposes (to track and analyse credit-related activity, assess borrower risk and detect early warning signs of default or financial distress) after credit is granted should not be classified as high-risk under point 5(b) of Annex III.

Evaluation of a collateral

• AI systems intended to evaluate collateral, so long as the AI system is intended to be used solely in relation to the collateral asset, for instance by assessing its features, its risk factors and/or its feasibility to be sold.

Providing credit or extended margin for leveraged trading products

• An AI system may be intended to be used by financial intermediaries to evaluate the extension of margin credit (whereby clients borrow capital against their existing securities) to existing or prospect clients. Despite being capable of determining access to financial resources to natural persons, such AI systems should not be classified as high-risk under point 5(b) of Annex III, if they are solely intended to be used for providing or extending margin credit as a non-essential private service.

Risk assessment and pricing

• An AI system intended to be used by an insurer that reviews applications for life insurance, so long as it qualifies as a risk assessment as defined above. This is usually the case if the AI system processes data provided by the applicant, such as age, health status, family history, lifestyle habits and occupation and relies on mortality tables, that estimate the probability of death for each applicant within a given period. This processing guides whether the insurer accepts the application and what conditions apply.
• An AI system intended to be used by an insurer to predict the expected annual medical cost for a risk group. The AI system may calculate and add expenses for administration and a profit margin. The final figure becomes the premium charged to members of that group.

Claims management 

• AI systems intended to be used for claims management in case of health insurance products in case the insured event happens. This is because these systems are intended to to verify whether a claim is valid under the policy terms or to determine the amount to be paid, they do not fall within the use case of point 5(c) of Annex III. These use cases are distinct from carrying out a risk assessment or pricing, even though their outcome might affect the enjoyment of the health insurance. In the same vein, those systems may fall outside the use case of point 5(c) of Annex III if they are intended to be used in claims management to improve the quality of claims processing and decision-making by contributing to the accuracy, by inter alia the validation of claimant information against various databases.

Product design for life insurance

• AI systems intended to be used in product design for life insurance fall outside the use case of point 5(c) of Annex III, if they are not intended to be used for risk assessment and pricing in individual cases. Those AI systems typically support a data-driven approach to product design by analysing large volumes of demographic, behavioural, and socio-economic data. By identifying emerging customer needs, risk patterns, and coverage gaps, the AI system enables insurers to create tailored life insurance products that are both relevant and competitively priced. For example, these AI systems can detect underserved segments and recommend product features or pricing models suited to their profiles. Scenario simulations and predictive modelling may help to assess the potential impact of new product offerings before launch, reducing development time and market risk. Those AI systems usually represent a prior and separate stage in the value chain of insurance companies. The product design may set the framework on what insurance products are offered and within which pricing framework. However, there is no evaluation of a natural person’s risk profile involved to determine whether to offer, decline, revoke or deliver services related to health and life insurance. 

Analysing calls and prioritising interventions

• An AI system intended to be used in emergency response centres where 112 calls are classified to assign the level of urgency and to route responders, using natural language processing. Such an AI system is high-risk, since the system is intended to evaluate and classify the calls, and to establish prioritisation in the dispatching of emergency first response services, including the actual dispatching of the services.
• An AI system is used to analyse emergency calls received by the police and to prioritise interventions according to the urgency and seriousness of the situation through the extraction of relevant information, such as relevant key words. The system assesses the urgency of the situation and the response to be given. Such an AI system is as high-risk, since it is directly involved in the evaluation, prioritisation and dispatching of incoming calls made by natural persons, and the system assesses which response should be given.

Supportive tools for human review

• An AI system intended to provide real-time decision support for emergency first response services call-talkers, effectively enhances human decision-making skills by acting as an assistant to the call-taker, listening for signs or signals of life-threatening emergencies in what the caller is describing. The AI system uses real-time automatic speech recognition technology, creating high-quality transcripts and during the call, the data is analysed and compared with historical data collected from previous emergency calls. Such an AI system is high-risk, since it evaluates the call made by a natural person by assisting the emergency first response services call-taker by helping to classify the call on the basis of the severity of the situation through the listening for keywords that might indicate a medical emergency.
• AI systems intended to determine the nature and severity of emergency situations in incoming calls that are put in a queue before they are picked up. The AI system’s role is to identify the emergency as life-threatening and alert the human call-taker if the waiting incoming call needs to be prioritized, while the decision to perform an automated triage or prioritise dispatch remains with the human call-taker. Such an AI system is high-risk, since it evaluates and classifies emergency calls into those that are life-threatening and those that are not.

Emergency triage systems

• AI systems used in emergency departments that are intended to prioritise patients, without performing clinical assessment medical acts or diagnostic functions . Such an AI system is high-risk, since it is intended to be used to establish priority in the dispatching of emergency healthcare patient triage systems.
• AI systems intended to be used as chatbots to trigger emergency medical emergency services in medical institutions without performing clinical assessment medical acts or diagnostic functions.

Rapid alert systems

• An AI system intended to be used to support decisions on allocating resources and pre-positioning assets to combat wildfires, where the system not only predicts the likelihood and the direction of wildfire evolution across an area’s protected forests, but also automatically triggers the dispatching of emergency first response services or provides clear recommendation of staff allocation in order to contain the wildfire. The system does not replace human decision-making but supports staff. Such an AI system is high-risk, since it is specifically intended to trigger emergency first response services, as well as to decide when and/or where to dispatch emergency first response services. The impact of such system is related to its decisive role for the dispatching of resources and to locations, which are decisions made in critical situations for the life and health of persons and the protection of their property.

Language tools

• AI systems intended to be used as a mental health crisis triage chatbot, which uses natural language processing to assess severity and urgency of chat-based contacts with emergency mental health services without performing clinical assessment or diagnostic functions. Such an AI system is high-risk, since it directs mental health intervention responses, carries a risk of misclassification and provides a first emergency triage function. 

Transcribing emergency calls

• An AI system intended to transcribe poor-quality emergency calls, or calls where callers may be panicked or unable to communicate, identifying relevant keywords and transcribing them automatically, thereby saving time that the call-taker would otherwise spend asking the caller to repeat themselves. Such an AI system falls outside the use case listed in point 5(d), since it aids the emergency first response services call-taker in the evaluation of a call made by a natural person by focusing on specific keywords related to emergency first response services, but does not evaluate or classify the call.

Forecasting

• AI systems that analyse various sets of data for fast indications of potential disasters before emergency first response services are made aware through emergency communications. These systems can for example identify flood-prone regions and suggest mitigation strategies, model potential wildfire spread patterns based on wind, temperature and vegetation density, process social media feeds to detect when a new topic is trending and monitor weather conditions and seismic activity. Such an AI system falls outside the use case listed in point 5(d), since it is not intended to decide when and/or where to dispatch emergency first response services, but rather resembles systems that are intended for the general forecasting of potential disasters, without a link to dispatching emergency first response services. 

Enhanced learning and training

• An AI driven simulation platform used to provide realistic training environments. The system creates among others scenario-based training for first responders, allowing them to practice responses to floods, wildfires and cyberattacks, followed by automated debriefings and response recommendations. Such an AI system falls outside the use case listed in point 5(d), since it is intended to provide a learning environment with realistic scenarios. In practice, such systems are not used to evaluate and classify incoming real-time calls from natural persons, nor do they actually dispatch or prioritise emergency first response services. The experience and information gained within this environment is valuable for the development of an AI system which is to be used for the intended purposes listed in point 5(d) of Annex III of the AI Act, but is not considered high-risk for learning purposes.

Identification of patients

• AI systems intended to be used to securely identify patients undergoing medical procedures, using biometric parameters. Such an AI system falls outside the use case listed in point 5(d), since the system is not intended to be use for one of the intended purposes listed therein.

AI systems related to the connectivity of the service

• AI systems intended to convey traffic on the network of received emergency calls or the lack of connectivity. Such an AI system falls outside the use case listed in point 5(d), since it is not inherently responsible for triggering harmful outcomes, even if it may affect the availability of services.

Estimated healing time

• AI systems intended to estimate the healing time for room- and bed-management in a hospital. Such an AI system falls outside the use case listed in point 5(d), since it is not intended to be use for one of the intended purposes lists therein.

Medical appointments

• AI systems intended to schedule medical appointments. Such an AI system falls outside the use case listed in point 5(d), since it is not intended to be use for one of the intended purposes listed therein and it does not relate to life-threatening emergencies.

Preparatory tools for prioritization

• An AI system intended to interpret weather risk data, such as of wildfire or flood, in advance of a probable emergency and to provide preparatory intelligence for the prioritisation of resources, such as pre-positioned firefighters or flood prevention devices, and their allocation pre-emptively. Such an AI system falls within the use case listed in point 5(d), since it is intended to be used for  establishing priority in the dispatching of emergency services. However, since the system performs a preparatory task to an assessment relevant for the prioritisation that will be evaluated and confirmed by human analysts and the competent commander, it falls under the exception in Article 6(3)(d) AI Act.

Supportive tools for human review

• AI systems intended to provide situational awareness of an emergency site to first responders, such as of where patients and casualties are located and what is e.g., their body temperature. This could be a system held by the on-site commander that gathers and combines the respective data from human first responders, from autonomous rescue robots or drones, and other sensor data (e.g., radioactivity or toxins). Such an AI system falls within the use case of point 5(d) of Annex III, since it is linked to emergency healthcare patient triage systems. However, since the system merely provides existing and factual information in a structured manner, without evaluating this information for the prioritisation or triage of the patients and casualties, it falls under the exceptions for AI systems intended to perform narrow procedural and preparatory tasks in Article 6(3)(a) and (d) AI Act.