AI Act classification memo: vendor credit-scoring model, Lumen Bankas UAB
Regulation: Regulation (EU) 2024/1689 (EU AI Act) · CELEX: 32024R1689 · Corpus data_as_of: 2026-09-02T09:05:48Z (EUR-Lex consolidated text, held by Rekvira) Prepared: 2026-09-23 · Status: Sources, not advice. The classification and the compliance conclusion are for the compliance lead.
Checkpoints (recorded without a person present):
- Checkpoint: the system is a vendor-built model that scores the creditworthiness of individuals applying for consumer loans of up to EUR 15,000. Loan officers see the score and three reason codes, and a loan officer makes the final decision. Lumen Bankas does not change the model and uses it only in internal tools. I treat Lumen Bankas as the deployer and the vendor as the provider.
1. System and intended purpose
- What it does: it produces a creditworthiness score, with three reason codes, for individuals who apply for consumer loans of up to EUR 15,000.
- Who it affects: individuals who apply for consumer credit at Lumen Bankas UAB, a specialised bank licensed in Lithuania, based in Vilnius, with 140 staff.
- Who built it: an outside vendor.
- Who runs it: Lumen Bankas. Its loan officers use the output, and a loan officer makes the final decision.
- Changes by the bank: none. The system is used under the bank's name only inside internal tools.
- Intended purpose is defined by the provider's documentation (Article 3(12)): "the use for which an AI system is intended by the provider … as specified in the information supplied by the provider in the instructions for use". The vendor's instructions for use were not reviewed for this memo. See §6.
2. Role
- Deployer. Article 3(4) defines a deployer as "a natural or legal person, public authority, agency or other body using an AI system under its authority except where the AI system is used in the course of a personal non-professional activity".
- Provider (the vendor). Article 3(3) defines a provider as a body "that develops an AI system … or that has an AI system … developed and places it on the market or puts the AI system into service under its own name or trademark".
- When a deployer becomes a provider. Article 25(1) makes a deployer a provider, subject to Article 16, if:
- (a) "they put their name or trademark on a high-risk AI system already placed on the market or put into service"; or
- (b) they make "a substantial modification"; or
- (c) they change the intended purpose.
On the facts given, the bank makes no modification and does not change the purpose. Point (a) is the one to check, because the bank uses the system "under its own name" in internal tools (see §6).
3. Classification under Article 6 and Annex III
Prohibited practices (Article 5(1)). Article 5(1)(c) prohibits social scoring that leads to "detrimental or unfavourable treatment … in social contexts that are unrelated to the contexts in which the data was originally generated or collected", or treatment that is "unjustified or disproportionate".
- Nothing in the facts points to this. The system scores credit applicants for credit.
- The inputs to the model were not reviewed, so this route has not been ruled out. The other points of Article 5(1) (manipulation, criminal-risk prediction, biometrics, emotion recognition) do not match the use described.
The high-risk test.
- Article 6(1), product safety route: this covers systems that are a safety component of, or are themselves, a product "covered by the Union harmonisation legislation listed in Annex I". Nothing in the facts suggests this applies.
- Article 6(2), Annex III route: "AI systems referred to in Annex III shall be considered to be high-risk."
- Annex III, point 5(b): "AI systems intended to be used to evaluate the creditworthiness of natural persons or establish their credit score, with the exception of AI systems used for the purpose of detecting financial fraud."
- Recital (58) (a recital, so not binding, but it explains the rule): "AI systems used to evaluate the credit score or creditworthiness of natural persons should be classified as high-risk AI systems, since they determine those persons' access to financial resources or essential services".
The Article 6(3) derogation. An Annex III system "shall not be considered to be high-risk where it does not pose a significant risk of harm … including by not materially influencing the outcome of decision making". At least one of these conditions must be met:
- (a) a "narrow procedural task";
- (b) improving "the result of a previously completed human activity";
- (c) detecting decision-making patterns "not meant to replace or influence the previously completed human assessment";
- (d) "a preparatory task to an assessment".
However, the same paragraph also says: "Notwithstanding the first subparagraph, an AI system referred to in Annex III shall always be considered to be high-risk where the AI system performs profiling of natural persons."
- Article 3(52) defines profiling by reference to GDPR Article 4(4).
- GDPR Article 4(4) defines it as automated processing "to evaluate certain personal aspects relating to a natural person, in particular to analyse or predict aspects concerning that natural person's … economic situation … reliability".
- Who may rely on the derogation: under Article 6(4), "A provider who considers that an AI system referred to in Annex III is not high-risk shall document its assessment". That assessment is the vendor's to make. The deployer does not make it.
What points each way
| Points toward high-risk | Points against, or open |
|---|---|
| The use matches the wording of Annex III 5(b) ("evaluate the creditworthiness of natural persons") | The fraud-detection exception in 5(b) could apply only if the model's intended purpose is fraud detection. Nothing in the facts says so |
| Recital (58) names credit scoring expressly | A loan officer makes the final decision. That is relevant to "materially influencing" under 6(3), but see the profiling rule |
| The 6(3) profiling rule: a credit score for an individual appears to fit "economic situation … reliability" in GDPR Art 4(4) | The Article 6(3) conditions (a)–(d) are framed around the system's intended purpose. This needs the vendor's documentation |
| Loan officers see the score and reason codes during the decision. The system is not described as an after-the-fact check | Any Article 6(4) assessment by the vendor was not seen |
When the rules apply. Under Article 113, the Regulation "shall apply from 2 August 2026". Article 6(1) obligations apply from 2 August 2027. On the held text, the Annex III deployer duties apply on the date the system is put into service. The corpus is EUR-Lex consolidated text as of 2026-09-02. Check separately whether any amending act has changed these dates (see §6).
4. Obligations that follow if the system is high-risk and the bank is the deployer
Found through lookup_obligations (deployer, high-risk), which returned Article 26 only, plus Article 13 as a related article. That index lists only articles whose titles name the role, so it is not the full set of duties. Articles 4, 27 and 86 below were found by search and reading.
- Article 4, AI literacy: "shall take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff".
- Article 26(1), use per the instructions: "take appropriate technical and organisational measures to ensure they use such systems in accordance with the instructions for use".
- Article 26(2), human oversight: "assign human oversight to natural persons who have the necessary competence, training and authority, as well as the necessary support".
- Article 26(3): paragraphs 1 and 2 apply "without prejudice to other deployer obligations under Union or national law".
- Article 26(4), input data: "to the extent the deployer exercises control over the input data, that deployer shall ensure that input data is relevant and sufficiently representative".
- Article 26(5), monitoring and incidents:
- The bank must monitor the system under the instructions for use.
- If it has reason to consider there is an Article 79(1) risk, it must inform the provider and the market surveillance authority and "suspend the use".
- It must report serious incidents "immediately", to the provider first.
- For financial institutions, "the monitoring obligation … shall be deemed to be fulfilled by complying with the rules on internal governance arrangements" under Union financial services law.
- Article 26(6), logs:
- Logs under the bank's control must be kept for "at least six months".
- Financial institutions "shall maintain the logs as part of the documentation kept pursuant to the relevant Union financial service law".
- Article 26(7), workers: "deployers who are employers shall inform workers' representatives and the affected workers" before putting a high-risk system into service at the workplace.
- Article 26(9), DPIA: the bank uses the provider's Article 13 information "to comply with their obligation to carry out a data protection impact assessment under Article 35 of Regulation (EU) 2016/679".
- Article 26(11), informing applicants: deployers of Annex III systems that make or help make decisions about natural persons "shall inform the natural persons that they are subject to the use of the high-risk AI system".
- Article 26(12): "Deployers shall cooperate with the relevant competent authorities".
- Article 27(1), fundamental rights impact assessment (FRIA): "deployers of high-risk AI systems referred to in points 5 (b) and (c) of Annex III, shall perform an assessment of the impact on fundamental rights". The contents are listed in (a)–(f).
- Article 27(2): this "applies to the first use".
- Article 27(3): "the deployer shall notify the market surveillance authority of its results".
- Article 27(4): the FRIA "shall complement that data protection impact assessment".
- Article 86(1), explanation on request: affected persons have "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".
- Article 86(3): this applies "only to the extent that the right … is not otherwise provided for under Union law".
- Article 13(3), what the vendor must supply: the instructions for use must include, among other things, performance limits, the "human oversight measures referred to in Article 14", and log mechanisms. These are the inputs for 26(1), 26(5), 26(9) and 27(1)(d)–(e).
Read but not applicable on these facts:
- Article 26(8) and Article 49(3): registration by deployers that are "public authorities, Union institutions, bodies, offices or agencies or persons acting on their behalf". The bank is a private entity.
- Article 26(10): post-remote biometric identification.
- Article 16: provider obligations. Relevant only if Article 25(1) applies. It was verified to exist but not read in full.
- Article 50: transparency duties for certain AI systems. Not read, because the system is not described as interacting directly with applicants.
5. Citations checked (verify_citation, corpus data_as_of 2026-09-02)
| Pinpoint | Exists | Quote |
|---|---|---|
| Art 3(3) AI Act (the tool labels it "Article 3, amendment (3)") | held | found |
| Art 3(4) (tool label "amendment (4)") | held | found |
| Art 3(52) (tool label "amendment (52)") | held | found |
| Art 3(12) | read in the full Art 3 text, not run through verify_citation separately | — |
| Art 4 | held | found |
| Art 5(1) | held | found |
| Art 6(1), 6(2), 6(3), 6(4) | held | found (each) |
| Annex III point 5(b) (tool resolved it at annex level, "ANNEX III") | held | found |
| Recital (58) | held | found |
| Art 13(3) | held | found |
| Art 16 | held | no quote checked |
| Art 25(1) | held | found |
| Art 26(1), (2), (3), (4), (5), (6), (7), (9), (11), (12) | held | found (each) |
| Art 27(1), (2), (3), (4) | held | found (each) |
| Art 49(3) | held | found |
| Art 86(1), 86(3) | held | found |
| Art 113 | held | found |
| GDPR Art 4(4) (tool label "Article 4, amendment (4)") | held | found |
No pinpoint failed. The only flag is Article 3(12): it was quoted from the full article text but not checked separately.
6. Open points for your decision
- Classification. Does Annex III 5(b) cover this system, and does the Article 6(3) profiling rule stop the derogation from applying? That is your call. Ask the vendor whether it has made an Article 6(4) non-high-risk assessment and registered under Article 49(2).
- "Under its own name" and Article 25(1)(a). Does internal branding amount to putting the bank's "name or trademark on a high-risk AI system"? If it does, the bank carries the Article 16 provider duties. Get the contract and the user interface screens and decide.
- Vendor documents needed:
- the instructions for use (Article 13(3));
- the intended purpose statement, including whether it excludes fraud detection;
- what logging the system does and who controls the logs;
- the inputs used (relevant to Article 5(1)(c) and Article 26(4)).
- FRIA and DPIA: plan the Article 27 FRIA to sit alongside the GDPR Article 35 DPIA. Confirm which Lithuanian authority receives the Article 27(3) notification.
- Financial-services carve-outs in 26(5) and 26(6): decide which Union financial services governance rules the bank relies on. Those rules are not in this corpus. DORA is loaded but was not searched for this memo.
- Dates: confirm that no amending act published after the 2026-09-02 corpus date changes the Article 113 dates for Annex III systems.
- Not held or not reviewed:
- Lithuanian implementing law and the designation of the market surveillance authority;
- Commission guidelines under Article 6(5);
- harmonised standards;
- the AI Office FRIA template (Article 27(5));
- EBA and national banking guidance;
- consumer credit law.
Not assessed at all: the model's accuracy, bias or data quality.
No assessment verdicts were recorded in Rekvira, because no officer has stated an applicability decision yet.