AI in the Finance Function Is Not High-Risk Under the EU AI Act

"AI in finance is high-risk under the EU AI Act" is one of the most repeated claims about the regulation, and it is wrong. No point of Annex III covers the finance function. What is caught is narrow, specific, and mostly sits in HR and in consumer lending rather than in accounting.

What Annex III actually lists

Annex III sets out the systems treated as high-risk under Article 6(2). Two of its points touch the areas a finance leader is likely to be asked about.

Point 4 covers employment, workers' management and access to self-employment, in two sub-points. The first covers AI systems intended to be used for the recruitment or selection of natural persons, extending to placing targeted job advertisements, analysing and filtering applications, and evaluating candidates. The second covers AI systems intended to be used to make decisions affecting terms of work-related relationships, extending to decisions on promotion or termination, task allocation based on individual behaviour or personal traits, and monitoring and evaluating performance and behaviour.

Point 5 covers essential private and public services. Two of its sub-points matter here. One covers AI systems intended to evaluate the creditworthiness of natural persons, with an express exception for systems used for the purpose of detecting financial fraud. The other covers AI systems intended to be used for risk assessment and pricing in relation to natural persons, and is limited to life and health insurance.

Two further points are worth knowing. Point 1 covers emotion recognition and biometric categorisation, though biometric verification, meaning confirming that a person is who they claim to be, is excluded. Point 3 covers education and vocational training, including evaluating learning outcomes and monitoring prohibited behaviour during tests, which can catch a firm running assessed internal training.

What is not there

This is the part worth stating plainly, because almost nobody does. There is no high-risk classification in Annex III for AI used in financial reporting, the month-end close, consolidation, accounts payable or receivable, reconciliations, treasury, cash forecasting, financial planning and analysis, management reporting, tax compliance, internal audit or external audit.

A finance function's realistic exposure to the high-risk regime is as a deployer of somebody else's HR tool or credit tool, not through its own accounting work. That distinction changes what a governance exercise should look for: the question is not "where are we using AI in finance" but "are we deploying a system that makes decisions about individual people, their employment, their credit or their insurance".

Two limits on the creditworthiness point

Creditworthiness assessment is high-risk, and there is a carve-out. Both halves of that sentence get mangled, and there are two further limits that are routinely missed.

The first is the words "natural persons". On the face of the provision, assessing the creditworthiness of corporate counterparties is not within it. For a finance team running B2B credit control, trade credit limits or customer onboarding for corporate customers, that is a significant distinction and it is almost always glossed over.

The second is the derogation in Article 6(3). A system listed in Annex III is not treated as high-risk where it does not pose a significant risk of harm to health, safety or fundamental rights, including by not materially influencing the outcome of decision making. That applies where the system performs a narrow procedural task, or improves the result of a previously completed human activity, or detects decision-making patterns or deviations from prior patterns without being meant to replace or influence the previous human assessment without proper human review, or performs a preparatory task to an assessment.

And then the proviso that defeats most credit-scoring arguments: notwithstanding that derogation, a system listed in Annex III is always high-risk where it performs profiling of natural persons. Most consumer credit scoring profiles natural persons, so the derogation rarely rescues it.

Anyone relying on the derogation also picks up a documentation duty. Under Article 6(4), a provider who considers that an Annex III system is not high-risk must document that assessment before the system is placed on the market or put into service, is subject to the registration obligation in Article 49(2), and must provide the documentation to national competent authorities on request. A conclusion reached informally is not a defence.

Why credit scoring is regulated rather than prohibited

A recurring client question is where lawful credit assessment ends and prohibited social scoring begins. Article 5(1)(c) prohibits AI systems that evaluate or classify people over a period of time based on their social behaviour or known, inferred or predicted personal or personality characteristics, where the resulting social score leads to detrimental or unfavourable treatment either in social contexts unrelated to the contexts in which the data was originally generated or collected, or that is unjustified or disproportionate to the social behaviour or its gravity.

The two limbs are the test: context-unrelatedness, and disproportionality. Conventional credit assessment on financial data, used to make credit decisions, fails neither limb. That is precisely why creditworthiness sits in Annex III as high-risk and regulated rather than in Article 5 as prohibited. The prohibitions have applied since 2 February 2025 and carry the top penalty tier of up to 35 million euro or 7% of worldwide turnover, so the distinction is not academic.

What a deployer actually has to do, and when

If a finance or HR function does deploy an Annex III high-risk system, Article 26 sets out its duties. Take appropriate technical and organisational measures to use the system in accordance with its instructions for use. Assign human oversight to natural persons who have the necessary competence, training and authority, and give them the support they need. Monitor operation on the basis of the instructions, and on identifying a serious incident inform the provider first, then the importer or distributor and the relevant market surveillance authority. Keep the logs the system generates automatically for at least six months unless other Union or national law provides otherwise. Use the information the provider supplies under Article 13 to comply with any data protection impact assessment obligation.

One duty deserves separate mention because it catches employers specifically: before putting a high-risk system into service at the workplace, deployers who are employers must inform workers' representatives and the affected workers.

A fundamental rights impact assessment is required under Article 27 for the creditworthiness and life and health insurance pricing points, and the omnibus regulation amended that article so an assessment may cross-reference a data protection impact assessment.

These duties do not apply yet. Annex III high-risk obligations now commence on 2 December 2027, as set out in our piece on the high-risk rules moving to December 2027. What does apply today is the transparency regime in Article 50 and the AI literacy duty, which we cover in what Article 4 actually requires.

How to run the assessment

Inventory systems by what decision they influence and about whom, not by which department uses them. Separate provider from deployer for each system, because the obligations differ. Record the Annex III analysis, including the reasons a system is considered outside it, since Article 6(4) requires that documentation of anyone relying on the derogation. And distinguish corporate from consumer exposure, because the creditworthiness point is confined to natural persons.

That is recognisably the same discipline as risk-based internal audit planning, where the rigour lies in documenting why something was scoped out, a point our analysis of the IIA Standards on design versus operating effectiveness develops. The authoritative text is the consolidated regulation, published at EUR-Lex; the European Commission's explanatory pages are useful but carry an express disclaimer that they are not legally binding.