knowledge 9 June 2026 |

Essential Services Spotlight: EU AI Act - Draft Guidelines on High-Risk AI Classification

As noted in our previous briefings (Employment Spotlight: EU AI Act - Draft Guidelines on High-Risk AI Classification and Critical Infrastructure Spotlight: EU AI Act - Draft Guidelines on High-Risk AI Classification), the European Commission's five-week public consultation on the draft Guidelines runs until 23 June 2026.

WHAT DO SERVICE PROVIDERS NEED TO KNOW?

As mentioned in our earlier briefings, the draft Guidelines confirm that the application date for the obligations relating to Annex III high-risk AI systems, including those concerning essential services, has been postponed from 2 August 2026 to 2 December 2027.  Such high-risk AI systems in use prior to 2 December 2027 are not subject to the AI Act unless they are subject to “significant changes in their design”, or if they are used by public authorities (in which case they must comply by 2 August 2030).  Otherwise, the substantive compliance obligations have not changed. Organisations deploying AI in the delivery of essential services should use this extended timeline to prepare and may also wish to submit feedback to the Commission during the open consultation period.

KEY TAKEAWAYS FROM THE DRAFT GUIDELINES

Below, we highlight the key takeaways from the Guidelines relating to Point 5 of Annex III, which covers access to and enjoyment of essential private services and essential public services and benefits.

The scope covers four distinct use cases

Point 5 of Annex III lists four high-risk use cases:

 (a)  AI systems used by or on behalf of public authorities to evaluate eligibility for essential public assistance benefits and services, including healthcare, and to grant, reduce, revoke, or reclaim such benefits;

 (b)  AI systems used to evaluate the creditworthiness of natural persons or establish their credit score (except for fraud detection);

 (c)  AI systems used for risk assessment and pricing in relation to natural persons in the case of life and health insurance; and

 (d)  AI systems used to evaluate and classify emergency calls, or to dispatch or prioritise emergency first response services, including emergency healthcare patient triage systems.

Essential services

One question that arises is the relationship between the broad definition of essential services, as understood from Recital 58 of the EU AI Act, and the operative provisions that actually trigger high-risk classification. Recital 58 frames essential services as those necessary for people to fully participate in society or improve their standard of living, and this policy rationale clearly underpins the inclusion of Point 5 of Annex III. However, it is worth noting that the broad definitional language does not, of itself, give rise to high-risk classification. The overall effect would appear to be a targeted framework in which the broad concept of essential services explains why the area warrants regulatory attention, whilst the specific use cases enumerated in Annex III, together with the Article 6(3) filter mechanism, ensure that high-risk obligations attach only where the risk to fundamental rights is most acute.

Credit scoring: a fraud detection exception applies

The Guidelines state that  the “evaluation of creditworthiness” refers to the assessment of a natural person’s ability and willingness to fulfil contractual obligations to pay for services provided or credit granted.  A “credit score” refers to the creation and building of a representation of an individual’s creditworthiness.  AI systems used for creditworthiness evaluation or credit scoring are classified as high-risk, but an important exception exists for AI systems whose main intended use is to detect financial fraud.   For the exception to apply, fraud detection must be the system's main intended use, preceding all other purposes. This will be the case where the AI system is mainly used for pattern recognition and anomaly detection, rather than for assessing ability to repay a loan.  By contrast, AI systems used for anti-money laundering and countering the financing of terrorism are not covered by this exception.  Notably, the insurance use case under point 5(c) does not benefit from any fraud detection exception - an AI system used for risk assessment or pricing in the case of health and life insurance will be classified as high-risk even if it also has a fraud detection feature.

The Article 6(3) filter: a mixed picture

Systems that engage in profiling are categorically excluded from the filter, and many AI systems evaluating eligibility for public benefits or assessing creditworthiness will constitute profiling. However, the Guidelines identify a genuine set of filter-eligible systems. These include: chatbots that provide only existing factual information (without evaluating it for the purposes of eligibility) to a case handler; AI systems that summarise medical reports for case handler review; translation tools used to translate applications submitted in a foreign language; and speech-to-text systems that create conversation summaries where the case handler is present and able to check the output. In the emergency services context, AI systems that provide preparatory weather-risk intelligence for resource pre-positioning may also qualify, provided the human commander retains the final decision on dispatching.

Practical examples

The following is a high-level summary of examples from the Guidelines which indicate where the boundaries sit.

What's in (high-risk):

• AI systems used to decide on the grant of unemployment benefits in individual cases, including those that analyse eligibility and recommend outcomes to case handlers.
• AI systems used to prioritise allocation of social housing or home-care services based on applicant profiles.
• AI systems used to create a credit score for the purpose of supporting decisions on consumer credit or mortgages.
• AI systems used by insurers to review life insurance applications and process applicant health data to guide acceptance and policy terms.
• AI systems in emergency call centres that classify 112/999 calls to assign urgency levels and route responders.
• AI systems used in emergency departments that are intended to prioritise patients, without performing clinical assessment, medical acts or diagnostic functions.

What's out (not high-risk):

• AI systems that proactively identify persons in need of preventive care, where the system's output is not used to influence access to such care.
• AI systems used for customer support related to the assessment of their creditworthiness.
• AI systems used for claims management in health insurance, where the purpose is to verify claim validity rather than carry out a risk assessment.
• AI systems that transcribe emergency calls by identifying keywords, without evaluating or classifying the call.
• AI systems used to schedule medical appointments or estimate healing time for hospital bed management. 

How can McCann FitzGerald LLP help?

For further information or assistance, please reach out to one of the key contacts below, or your usual contact at McCann FitzGerald LLP.