AI system classification memo

An in-house putting into service is still a "putting into service" under the Act:

"'putting into service' means the supply of an AI system for first use directly to the deployer or for own use in the Union for its intended purpose;" (Article 3(11))

Working role: deployer. This depends on the Article 25(1)(a) point below.

"'deployer' means 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;" (Article 3(4))

The vendor is the provider on the facts given:

"'provider' means a natural or legal person, public authority, agency or other body that develops an AI system or a general-purpose AI model or that has an AI system or a general-purpose AI model developed and places it on the market or puts the AI system into service under its own name or trademark, whether for payment or free of charge;" (Article 3(3))

What points against the deployer role: the brief says the bank uses the system "under its own name". Article 25(1) moves provider status onto a deployer in these cases:

"Any distributor, importer, deployer or other third-party shall be considered to be a provider of a high-risk AI system for the purposes of this Regulation and shall be subject to the obligations of the provider under Article 16, in any of the following circumstances: (a) they put their name or trademark on a high-risk AI system already placed on the market or put into service, without prejudice to contractual arrangements stipulating that the obligations are otherwise allocated; (b) they make a substantial modification to a high-risk AI system … ; (c) they modify the intended purpose of an AI system …" (Article 25(1))

• Point (b) does not apply on the facts given ("does not change the model").
• Point (c) does not apply. The bank uses the system for its stated creditworthiness purpose.
• Point (a) is open. If the internal tool shows the model as a Lumen-branded system, the literal text could be engaged. If only the bank's workflow screens carry its name, and the vendor stays identified as the provider, the deployer reading is stronger. The proviso about "contractual arrangements" makes the vendor contract relevant.
• If Article 25(1) applied, the vendor "shall no longer be considered to be a provider of that specific AI system" (Article 25(2)), and the bank would take on the provider obligations under Article 16.

Checkpoint (recorded): this memo treats Lumen Bankas as the deployer. It assumes that "under its own name" means internal branding of the bank's own tools, not the bank's name or trademark on the AI system itself. See open point 1.

3.1 Prohibited practices (Article 5): screened, nothing indicated

The nearest candidate is social scoring in Article 5(1)(c): "the evaluation or classification of natural persons … based on their social behaviour or known, inferred or predicted personal or personality characteristics, with the social score leading to either or both of the following: (i) detrimental or unfavourable treatment … in social contexts that are unrelated to the contexts in which the data was originally generated or collected; (ii) detrimental or unfavourable treatment … that is unjustified or disproportionate …" (Article 5(1)).

The facts given show no unrelated-context use and no disproportionate treatment. Still, the vendor should confirm which data sources the model uses (open point 5).

3.2 Article 6(1), Annex I route: not engaged

A standalone credit-scoring model is not described as a safety component of an Annex I product, or as such a product itself (Article 6(1)).

3.3 Article 6(2) and Annex III: engaged on its face

"In addition to the high-risk AI systems referred to in paragraph 1, AI systems referred to in Annex III shall be considered to be high-risk." (Article 6(2))

Annex III, point 5 ("Access to and enjoyment of essential private services and essential public services and benefits"), (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;" (ANNEX III(5))

Recital (58): "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 such as housing, electricity, and telecommunication services."

Points towards high-risk:

• The stated purpose, scoring individual consumer-loan applicants, matches the wording of point 5(b).
• The fraud-detection exception does not match the stated purpose.
• Recital (58) also carves out systems "provided for by Union law … for prudential purposes to calculate credit institutions' … capital requirements". This model is used for origination decisions, not capital calculation.
• Nothing in Annex III(5)(b) ties the classification to loan size. The EUR 15,000 cap does not affect it.

3.4 Article 6(3) derogation: text makes it unavailable where profiling occurs

"By derogation from paragraph 2, an AI system referred to in Annex III shall not be considered to be high-risk 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." Conditions: "(a) … a narrow procedural task; (b) … improve the result of a previously completed human activity; (c) … detect decision-making patterns … ; or (d) … a preparatory task to an assessment relevant for the purposes of the use cases listed in Annex III." (Article 6(3))

"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 6(3))

"'profiling' means profiling as defined in Article 4, point (4), of Regulation (EU) 2016/679;" (Article 3(52)) GDPR: "'profiling' means any form of automated processing of personal data consisting of the use of personal data 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, behaviour …" (GDPR Article 4(4))

What points each way:

• For the derogation: a loan officer makes the final decision, so one could argue the score is a "preparatory task" under point (d).
• Against it: the model uses personal data to predict an applicant's economic situation and reliability. That is the GDPR Article 4(4) definition of profiling. If it applies, the final sentence of Article 6(3) makes the system high-risk "always", whatever the human-in-the-loop design. A score shown with reason codes to the decision-maker is also hard to square with "not materially influencing the outcome of decision making".
• Who makes the call: the derogation assessment belongs to the provider: "A provider who considers that an AI system referred to in Annex III is not high-risk shall document its assessment before that system is placed on the market or put into service." (Article 6(4)). A deployer cannot declare the system out of scope on its own.

3.5 Classification the text points to

The held text points to a high-risk AI system under Article 6(2) and Annex III, point 5(b), with the Article 6(3) derogation excluded by the profiling sentence. This memo does not make that call. The compliance lead records the verdict.

3.6 Timing (Article 113, as amended by 2026/1744)

"(a) Chapters I and II shall apply from 2 February 2025, with the exception of Article 5(1), first subparagraph, points (ba) and (bb), and Article 5(1a) and (1b) which shall apply from 2 December 2026; … (c) Chapter III, Sections 1, 2, and 3, with the exception of Article 6(5), shall apply from: (i) 2 December 2027 as regards AI systems classified as high-risk pursuant to Article 6(2) and Annex III; …" (Article 113)

My reading: Articles 4 and 5 (Chapters I and II) apply now. Articles 6 and 26–27 (Chapter III, Sections 1 and 3) apply to this system from 2 December 2027. The system goes live before that date. See open point 3 on transitional treatment of systems already in use. The Omnibus changed these dates. Earlier summaries that give 2 August 2026 for Annex III systems are out of date against this text.

The heading index (lookup_obligations, role=deployer, system_class=high-risk) returned only Article 26. The other items come from reading the articles directly. The index is not a complete duty set.

GDPR, alongside the AI Act (not exhaustive): GDPR Article 22(1) gives "the right not to be subject to a decision based solely on automated processing, including profiling". The loan-officer final decision matters here, and the human review needs to be real, not a rubber stamp. The DPIA duty (GDPR Article 35, referred to in AI Act Article 26(9)) was not read in this run.

Every pinpoint was run through verify_citation against consolidated text 02024R1689-20260727 (AI Act) or 32016R0679 (GDPR), with the quoted wording, on 2026-09-24. All returned exists: held, quote: found. None failed.

Cited but not verified by quote: AI Act Art 6(1) and Art 16, and GDPR Art 35. These are only referred to, not quoted.

Not held by the corpus: Lithuanian implementing law, including which authority is the market surveillance authority. Also not held: the Commission's Article 6(5) guidelines, harmonised standards, EBA or Bank of Lithuania guidance, and the consumer-credit directive.

1. Article 25(1)(a), name or trademark. Confirm how the system is labelled in the internal tools and what the vendor contract says about allocating provider obligations. If the bank's name is on the AI system itself, the provider role (Article 16) needs its own review.
2. Classification verdict. Decide whether to record the system as high-risk under Art 6(2) and Annex III(5)(b). Ask the vendor whether it has made any Art 6(3) derogation assessment under Art 6(4), and on what basis, given the profiling sentence.
3. Timing and transition. Under Art 113(c)(i), deployer duties for this system apply from 2 December 2027. Check the transitional rules for systems already in service before that date (Art 111, not read in this run). Decide whether to comply early. The FRIA is tied to "first use" (Art 27(2)).
4. Article 86 timing. Chapter IX appears to fall under the general application date, but Art 86 depends on the Annex III high-risk classification, which applies from 2 December 2027. Counsel should settle when the explanation right bites. Also settle how it relates to GDPR rights, given Art 86(3).
5. Vendor deliverables. Ask for the instructions for use and the Art 13 information needed for Art 26(1) and 26(9) and Art 27(1)(d). Ask for the model's input data sources (for the Art 5(1)(c) screen and Art 26(4)), the log access arrangements (Art 26(6)), and serious-incident contact points (Art 26(5)).
6. Market surveillance authority. Confirm the Lithuanian authority designated under Art 74(6) for the Art 27(3) FRIA notification. This is outside the corpus.
7. Scope of the financial-institution deeming rules. Confirm which CRD/CRR internal-governance rules the bank will rely on to meet Art 26(5) (monitoring only) and Art 26(6) (logs).

Assessments were not recorded with record_assessment. The tool is meant for use after the officer states applicability, and no officer was present in this run.