Responsible AI in Polish Higher-Education Administration under the EU AI Act: High-Risk Use Cases, Fundamental-Rights Impact Assessment, and Institutional Accountability

Probably not a substantial claim: may not pass S3 (load-bearing).

Comparable coordination problems recur across public-sector artificial intelligence adoption more broadly.

“Comparable coordination problems recur across public-sector artificialintelligence adoption more broadly (Zuiderwijk et al., 2021).”

• cites Zuiderwijk et al., 2021 sourceSupported

Probably not a substantial claim: may not pass S2 (checkable).

An accountability record is valuable only if it enables information, justification, and consequences rather than merely accumulating paperwork.

“In publicadministration terms, an accountability record is valuable only if it enables information, justification, and consequences rather than merely accumulating paperwork (Busuioc, 2021).”

• cites Busuioc, 2021 sourcePartly supported

  In “Accountability: A Conceptual Frame” (page unavailable), Busuioc says meaningful accountability requires information, explanation or justification, and the possibility of consequences; transparency alone is insufficient. The paper does not specifically assess the value of an accountability record or contrast it with accumulating paperwork, so the claim applies its framework to records rather than reporting a point the paper makes about them. (LLM, conf=0.80)

  Experimental — accuracy not yet measured

Classified as probably not a substantial claim.

A process-based account treats independent professional judgement as non-delegable rather than satisfied by nominal human presence.

“A processbased account similarly treats independent professional judgement as non-delegable rather than satisfied by nominal human presence (Shukla, 2026).”

• cites Shukla, 2026 sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing).

Work on trustworthy AI, standards, and risk informed the distinction between formal compliance and institutional capability.

“Work on trustworthy AI, standards, and risk informed the distinction between formal compliance and institutional capability(European Commission, 2019;Gamito & Marsden, 2024;Kusche, 2024;Laux et al., 2024).”

• cites European Commission, 2019Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Gamito & Marsden, 2024 sourceNot checked — no full text of the source
• cites Kusche, 2024 sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing).

FRIA scholarship and official fundamental-rights research informed the assessment architecture.

“FRIA scholarship and official fundamental-rights research informed the assessment architecture(Bertaina et al., 2025; European Union Agency for Fundamental Rights, 2025;Janssen et al., 2022;Mantelero, 2024;Thomaidou & Limniotis, 2025).”

• cites Bertaina et al., 2025 sourceNot checked — no full text of the source
• cites European Union Agency for Fundamental Rights, 2025Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Janssen et al., 2022 sourceSupported

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Rights-aware educational research establishes the relevance of learner choice, academic values, legal literacy, and context-specific governance.

“Rights-aware educational research establishes the relevance of learner choice, academic values, legal literacy, and context-specific governance(Berendt et al., 2020;Lukianova et al., 2025;Memarian & Doleck, 2023;Oncioiu & Bularca, 2025;Panagopoulou et al., 2025;Temper et al., 2025).”

• cites Berendt et al., 2020 sourcePartly supported

  The Introduction and §4 emphasize learner choice and agency, while §4 says approaches to AI regulation differ with national views of education and culture. §2.3 discusses the school as a protected space, which bears on academic values. The paper discusses legal rights and privacy rules, but the passages do not establish legal literacy as a distinct research or governance priority. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites Lukianova et al., 2025 sourcePartly supported

  The Conclusions (page not available) emphasize humanistic values, academic integrity, and educational policy at national and institutional levels; the Results distinguish the Polish and Ukrainian contexts. These support the academic-values and context-specific-governance parts of the claim. The paper discusses personalized learning, but does not establish the relevance of learner choice or legal literacy. (LLM, conf=0.86)

  Experimental — accuracy not yet measured
• cites Memarian & Doleck, 2023 sourceNot checked — no full text of the source

May combine two claims.

The duty of AI literacy is one of documented effort rather than of result, which increases rather than reduces the evidential burden.

“The duty is therefore one of documented effort rather than of result, which increases rather than reduces the evidential burden addressed in Section 4.4.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

When the Annex III regime applies, Article 26 requires deployers to follow instructions, assign competent, trained, and authorised human oversight, ensure relevance and representativeness of input data where under their control, monitor operation, report specified incidents, and retain logs under their control for an appropriate period of at least six months.

“When the Annex III regime applies, Article 26 requires deployers to follow instructions, assign competent, trained, and authorised human oversight, ensure relevance and representativeness of input data where under their control, monitor operation, report specified incidents, and retain logs under their control for an appropriate period of at least six months.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Current Commission guidance supports a role-and context-sensitive approach to AI literacy.

“Current Commission guidance supports a role-and context-sensitive approach to AI literacy (European Commission, 2026a).”

• cites European Commission, 2026aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

The AI Act does not settle whether such a fine may be imposed on a public university: Article 99(8) leaves each Member State to lay down rules on the extent to which administrative fines may be imposed on public authorities and bodies established in that State.

“The AI Act does not settle whether such a fine may be imposed on a public university: Article 99(8) leaves each Member State to lay down rules on the extent to which administrative fines may be imposed on public authorities and bodies established in that State.”

May combine two claims.

A university may reuse DPIA material inside a FRIA, but may not treat a completed DPIA as discharging Article 27.

“A university may therefore reuse DPIA material inside a FRIA, but may not treat a completed DPIA as discharging Article 27.”

May combine two claims.

Neither the label 'educational AI' nor a human signature at the end of a workflow determines the applicable legal regime.

“Yet neither the label 'educational AI' nor a human signature at the end of a workflow determines the applicable legal regime.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

In Poland the question of administrative fines on public universities falls to the Act of 3 July 2026, whose provisions on administrative fines apply from 28 October 2026.

“In Poland that question falls to the Act of 3 July 2026, whose provisions on administrative fines apply from 28 October 2026.”

Probably not a substantial claim: may not pass S4 (based).

Procurement documents cannot contract out of a statutory role under the AI Act.

“Procurement documents should allocate tasks but cannot contract out of a statutory role.”

Probably not a substantial claim: may not pass S2 (checkable).

Public universities require close attention because they are public-law bodies.

“Public universities require close attention because they are public-law bodies.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Transparency does not legitimise an otherwise prohibited or unlawful system; it addresses a separate risk.

“Transparency does not legitimise an otherwise prohibited or unlawful system; it addresses a separate risk.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A tool that checks whether a mandatory field is empty, without interpreting content or affecting substantive assessment, may perform a narrow procedural task.

“Conversely, a tool that checks whether a mandatory field is empty, without interpreting content or affecting substantive assessment, may perform a narrow procedural task.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Article 1(13) of Regulation (EU) 2026/1744 replaces paragraphs 4 and 5 of Article 27.

“Article 1(13) of Regulation (EU) 2026/1744 replaces paragraphs 4 and 5.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Bias can enter at multiple stages of the machine-learning pipeline and affect groups defined not only by race, gender, or nationality but also by socioeconomic status and disability.

“shows that bias can enter at multiple stages of the machine-learning pipeline and affect groups defined not only by race, gender, or nationality but also by socioeconomic status and disability (Baker & Hawn, 2022).”

• cites Baker & Hawn, 2022 sourceNot checked — no full text of the source

May combine two claims.

This amendment confirms rather than dissolves the boundary between FRIA and DPIA: it authorises evidential cross-reference between the two assessments while leaving their triggers, subjects and analytical objects distinct.

“This amendment confirms rather than dissolves the boundary defended in Section 4.2: it authorises evidential cross-reference between the two assessments while leaving their triggers, subjects and analytical objects distinct.”

Classified as probably not a substantial claim.

Intention inference should receive a separate legal analysis rather than being treated as automatically prohibited solely because it appears in the broader definition of emotion-recognition systems.

“Intention inference should therefore receive a separate legal analysis rather than being treated as automatically prohibited solely because it appears in the broader definition.”

May combine two claims.

Breach of Article 50 falls within Article 99(4)(g), which subjects it to administrative fines of up to EUR 15 000 000 or, where the offender is an undertaking, up to 3 % of its total worldwide annual turnover for the preceding financial year, whichever is higher; for SMEs, including start-ups, Article 99(6) caps the fine at whichever of the amount or the percentage is the lower.

“Breach of Article 50 falls within Article 99(4)(g), which subjects it to administrative fines of up to EUR 15 000 000 or, where the offender is an undertaking, up to 3 % of its total worldwide annual turnover for the preceding financial year, whichever is higher; for SMEs, including start-ups, Article 99(6) caps the fine at whichever of the amount or the percentage is the lower.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Article 1(7)(a) of the Omnibus inserts two new prohibitions into the first subparagraph of Article 5(1), as points (ba) and (bb): the first concerns AI systems generating or manipulating realistic intimate material depicting an identifiable natural person without that person's freely given, specific, informed, unambiguous and explicit consent; the second concerns material or performances within a

“Secondly, Article 1(7)(a) of the Omnibus inserts two new prohibitions into the first subparagraph of Article 5(1), as points (ba) and (bb): the first concerns AI systems generating or manipulating realistic intimate material depicting an identifiable natural person without that person's freely given, specific, informed, unambiguous and explicit consent; the second concerns material or performances”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

The rules in Sections 1-3 of Chapter III concerning Article 6(2) and Annex III systems apply from 2 December 2027; Article 6(1) and Annex I product-related rules apply from 2 August 2028.

“The rules in Sections 1-3 of Chapter III concerning Article 6(2) and Annex III systems apply from 2 December 2027; Article 6(1) and Annex I product-related rules apply from 2 August 2028.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Commentary on the Commission's guidelines observes that they address emotions without supplying any corresponding treatment of intentions, so that intention inference appears to remain outside the prohibition.

“Commentary on those guidelines observes that they address emotions without supplying any corresponding treatment of intentions, so that intention inference appears to remain outside the prohibition (Dushi, 2026).”

• cites Dushi, 2026Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The deployer must notify the market-surveillance authority after completing the FRIA and update the assessment where relevant factors change.

“The deployer must notify the market-surveillance authority after completing the assessment and update the assessment where relevant factors change.”

Classified as probably not a substantial claim. May combine two claims.

Article 3(39) defines an emotion-recognition system more broadly as identifying or inferring emotions or intentions based on biometric data, whereas Article 5(1)(f) refers to inferring emotions.

“Article 3(39) defines an emotion-recognition system more broadly as identifying or inferring emotions or intentions based on biometric data, whereas Article 5(1)(f) refers to inferring emotions.”

Probably not a substantial claim: may not pass S4 (based).

A university that uses an AI system under its authority will normally be a deployer under the AI Act.

“A university that uses an AI system under its authority will normally be a deployer.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Influence may arise because reviewers see only the AI-ranked shortlist, because the output anchors their judgement, because departure requires exceptional justification, or because time and workload make rejection unrealistic.

“Influence may arise because reviewers see only the AI-ranked shortlist, because the output anchors their judgement, because departure requires exceptional justification, or because time and workload make rejection unrealistic.”

May combine two claims.

The new prohibitions apply from December 2026 and are relevant to university disciplinary and safeguarding processes.

“The prohibitions apply from December 2026 and are relevant to university disciplinary and safeguarding processes.”

May combine two claims.

The provider must document any conclusion that a listed system falls outside the high-risk category, and the HEI deployer must verify that its local workflow remains consistent with that conclusion.

“The provider must document any conclusion that a listed system falls outside the high-risk category, and the HEI deployer must verify that its local workflow remains consistent with that conclusion.”

Probably not a substantial claim: may not pass S4 (based).

The correct unit of analysis for AI Act classification is the deployed workflow: input, output, interface, default, reviewer practice, legal consequence, and available remedy.

“The correct unit of analysis is the deployed workflow: input, output, interface, default, reviewer practice, legal consequence, and available remedy.”

May combine two claims.

Admissions, learning-outcome evaluation, level determination, test monitoring, and employment uses may fall within Annex III of the AI Act.

“The analysis shows that admissions, learning-outcome evaluation, level determination, test monitoring, and employment uses may fall within Annex III, whereas narrow procedural tools require fact-specific assessment; emotion inference in education is generally prohibited.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Responsible deployment requires meaningful human authority, intelligible reasons, contestability, accessibility, and auditable institutional ownership.

“Responsible deployment requires meaningful human authority, intelligible reasons, contestability, accessibility, and auditable institutional ownership.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Where an educational deployment infers intentions from biometric data, the safer institutional course is to treat the classification question as open, record the analysis, and apply Article 50(3) notification duties and the GDPR in full.

“Where an educational deployment infers intentions from biometric data, the safer institutional course is to treat the classification question as open, record the analysis, and apply Article 50(3) notification duties and the GDPR in full.”

Classified as probably not a substantial claim. May combine two claims.

The FRIA must describe the deployer's processes in which the system will be used, the intended period and frequency of use, categories of natural persons and groups likely to be affected, specific risks of harm to their fundamental rights, human-oversight measures, and measures to be taken if risks materialise, including internal governance and complaint mechanisms.

“The FRIA must describe the deployer's processes in which the system will be used, the intended period and frequency of use, categories of natural persons and groups likely to be affected, specific risks of harm to their fundamental rights, human-oversight measures, and measures to be taken if risks materialise, including internal governance and complaint mechanisms.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

The 2026 Digital Omnibus changed application dates while preserving the architecture of the AI Act.

“The 2026 Digital Omnibus changed application dates while preserving the architecture.”

Classified as probably not a substantial claim.

Annex III, point 3 of the AI Act identifies four education and vocational-training categories.

“Annex III, point 3 identifies four education and vocational-training categories.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Article 50 transparency obligations apply from 2 August 2026.

“Article 50 transparency obligations apply from 2 August 2026.”

May combine two claims.

Profiling precludes reliance on the Article 6(3) exclusion for a listed use; it does not independently bring an unlisted intended purpose into Annex III.

“Profiling therefore precludes reliance on the Article 6(3) exclusion for a listed use; it does not independently bring an unlisted intended purpose into Annex III.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The December 2027 deferral of Annex III obligations is an implementation runway, not a legal vacuum.

“The December 2027 deferral of Annex III obligations is an implementation runway, not a legal vacuum.”

Probably not a substantial claim: may not pass S4 (based).

For a system already referred to in Annex III, Article 6(3) prevents an indiscriminate conclusion that every deployment is high-risk.

“For a system already referred to in Annex III, Article 6(3) prevents an indiscriminate conclusion that every deployment is high-risk.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

The same institution may be a body governed by public law for the purposes of the fundamental rights impact assessment and, simultaneously, an undertaking, or indeed an SME, for the purposes of the fine ceiling.

“the same institution may be a body governed by public law for the purposes of the fundamental rights impact assessment and, simultaneously, an undertaking, or indeed an SME, for the purposes of the fine ceiling.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Article 4 AI literacy has applied since 2 February 2025, as has the Article 5(1)(f) prohibition relevant to this study.

“Article 4 AI literacy has applied since 2 February 2025, as has the Article 5(1)(f) prohibition relevant to this study.”

Classified as probably not a substantial claim. May combine two claims.

Article 27 permits reliance on previously conducted assessments or provider information where appropriate, but responsibility remains contextual: a vendor's generic fundamental-rights brochure cannot assess the applicant population, decision route, local data, language, accessibility, or remedy at a particular university.

“Article 27 permits reliance on previously conducted assessments or provider information where appropriate, but responsibility remains contextual: a vendor's generic fundamental-rights brochure cannot assess the applicant population, decision route, local data, language, accessibility, or remedy at a particular university.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

A recommendation that changes access to education, financial support, progression, employment, or a disciplinary process may affect a person as seriously as a formally automated decision.

“A recommendation that changes access to education, financial support, progression, employment, or a disciplinary process may affect a person as seriously as a formally automated decision.”

Classified as probably not a substantial claim. May combine two claims.

No single factor is conclusive for the contextual public-service test, and the analysis must be recorded for the particular deployment rather than assumed in either direction.

“No single factor is conclusive, and the analysis must be recorded for the particular deployment rather than assumed in either direction.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Article 27 does not prescribe a general six-week consultation period with equality bodies, consumer organisations, or data-protection authorities.

“First, Article 27 does not prescribe a general six-week consultation period with equality bodies, consumer organisations, or data-protection authorities.”

Probably not a substantial claim: may not pass S2 (checkable).

These obligations show why governance cannot be left to an individual lecturer or admissions officer.

“These obligations show why governance cannot be left to an individual lecturer or admissions officer.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In learning analytics, privacy and ethics are intertwined with trust, accountability, and transparency rather than confined to data security.

“In learning analytics, privacy and ethics are intertwined with trust, accountability, and transparency rather than confined to data security (Pardo & Siemens, 2014).”

• cites Pardo & Siemens, 2014 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Article 1(5) of Regulation (EU) 2026/1744 replaced Article 4 in its entirety: providers and deployers must now take measures to support the development of AI literacy, and the provision states expressly that it does not require them to guarantee any specific level of AI literacy in any individual.

“First, Article 1(5) of Regulation (EU) 2026/1744 replaced Article 4 in its entirety: providers and deployers must now take measures to support the development of AI literacy, and the provision states expressly that it does not require them to guarantee any specific level of AI literacy in any individual.”

Probably not a substantial claim: may not pass S4 (based).

Throughout the runway, currently applicable law continues to matter.

“Throughout the runway, currently applicable law continues to matter.”

May combine two claims.

These routes are not interchangeable, and an internal 'AI ethics committee' cannot displace statutory competences.

“These routes are not interchangeable, and an internal 'AI ethics committee' cannot displace statutory competences.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Learning-analytics scholarship has long identified informed consent, privacy, deidentification, data interpretation, and institutional data management as ethical governance questions rather than peripheral technical details.

“Learning-analytics scholarship has long identified informed consent, privacy, deidentification, data interpretation, and institutional data management as ethical governance questions rather than peripheral technical details (Slade & Prinsloo, 2013).”

• cites Slade & Prinsloo, 2013 sourceNot checked — no full text of the source

May combine two claims.

A controller cannot use trade secrecy as a blanket refusal; competing rights require a lawful balancing route.

“Nor can a controller use trade secrecy as a blanket refusal; competing rights require a lawful balancing route.”

Classified as probably not a substantial claim. May combine two claims.

A procurement specification determines which data can be exported, whether the institution receives meaningful documentation, whether a model update is notified, whether group-specific testing is possible, whether logs remain accessible, and whether an appeal can be reconstructed after termination.

“A procurement specification determines which data can be exported, whether the institution receives meaningful documentation, whether a model update is notified, whether group-specific testing is possible, whether logs remain accessible, and whether an appeal can be reconstructed after termination.”

May combine two claims.

Article 2(1)(4) of the Act disapplies the statute to scientific research and development within the meaning of Article 4(2) and (3) of the Law on Higher Education and Science, but only to the extent that the activity does not involve testing in real-world conditions within the meaning of Article 3(57) of Regulation (EU) 2024/1689, and Article 2(1)(5) covers systems put into service exclusively for

“Article 2(1)(4) of the Act disapplies the statute to scientific research and development within the meaning of Article 4(2) and (3) of the Law on Higher Education and Science, but only to the extent that the activity does not involve testing in real-world conditions within the meaning of Article 3(57) of Regulation (EU) 2024/1689, and Article 2(1)(5) covers systems put into service exclusively for”

Probably not a substantial claim: may not pass S2 (checkable).

Many governance failures are fixed before a system is ever used.

“Many governance failures are fixed before a system is ever used.”

• AssumptionRests on an assumption: Many governance failures are fixed before a system is ever used.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Faster rejection of incomplete applications is not a public value if the tool mistakes foreign documents and shifts correction costs to applicants.

“Faster rejection of incomplete applications is not a public value if the tool mistakes foreign documents and shifts correction costs to applicants.”

May combine two claims.

The European Data Protection Board's guidance on automated decision-making and DPIA reinforces the need for a functional Article 22 analysis, intelligible information, prior risk assessment, and safeguards that operate in the real processing context.

“The European Data Protection Board's guidance on automated decision-making and DPIA reinforces the need for a functional Article 22 analysis, intelligible information, prior risk assessment, and safeguards that operate in the real processing context (European Data Protection Board, 2018a, 2018b).”

• cites European Data Protection Board, 2018a, 2018bNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

Article 22 gives a person the right not to be subject to a decision based solely on automated processing, including profiling, that produces legal effects or similarly significantly affects that person, subject to limited exceptions and safeguards.

“Article 22 gives a person the right not to be subject to a decision based solely on automated processing, including profiling, that produces legal effects or similarly significantly affects that person, subject to limited exceptions and safeguards.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

A high override rate can indicate healthy review, poor model performance, unclear criteria, or a mismatch between product and context.

“A high rate can indicate healthy review, poor model performance, unclear criteria, or a mismatch between product and context.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Lowest-price or convenience purchasing can externalise long-term rights and remedy costs.

“Lowest-price or convenience purchasing can externalise long-term rights and remedy costs.”

May combine two claims.

Admissions, learning-outcome assessment, determination of educational level, test monitoring, and employment uses may enter the AI Act's Annex III high-risk regime.

“Admissions, learning-outcome assessment, determination of educational level, test monitoring, and employment uses may enter the AI Act's Annex III highrisk regime.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Responsible AI in higher-education administration begins with correct classification but cannot end there.

“Responsible AI in higher-education administration begins with correct classification but cannot end there.”

Probably not a substantial claim: may not pass S4 (based).

When a model leaves a genuinely covered research activity and is used to rank actual applicants, determine student opportunities, or manage employees, the exclusion ceases to apply, and the full regime attaches.

“when a model leaves a genuinely covered research activity and is used to rank actual applicants, determine student opportunities, or manage employees, the exclusion ceases to apply, and the full regime attaches.”

Probably not a substantial claim: may not pass S2 (checkable).

Cross-reference between FRIA and DPIA should reduce duplication, not erase the broader rights analysis.

“Crossreference should reduce duplication, not erase the broader rights analysis.”

Classified as probably not a substantial claim.

Meaningful oversight has at least six elements.

“Meaningful oversight has at least six elements.”

Probably not a substantial claim: may not pass S2 (checkable).

A vendor cannot provide a useful explanation merely by supplying hundreds of feature weights without explaining, in intelligible terms, the procedure and principles applied.

“A vendor cannot provide a useful explanation merely by supplying hundreds of feature weights without explaining, in intelligible terms, the procedure and principles applied”

Classified as probably not a substantial claim.

Notice, justification, and redress conditions are core elements of algorithmic legitimacy in institutional governance.

“These notice, justification, and redress conditions are also core elements of algorithmic legitimacy in institutional governance (Iwanowska, 2025).”

• cites Iwanowska, 2025 sourceNot checked — no full text of the source

May combine two claims.

FRIA and DPIA both protect people and can share evidence, but they ask different legal questions.

“FRIA and DPIA both protect people and can share evidence, but they ask different legal questions.”

May combine two claims.

A non-public HEI using the same system must apply the contextual public-service test and record its conclusion; regardless of the FRIA result, it may still need a DPIA.

“A non-public HEI using the same system must apply the contextual public-service test proposed above and record its conclusion; regardless of the FRIA result, it may still need a DPIA.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

A technical explanation describes how a system produced or contributed to an output; a legal or academic reason explains why the authorised body reached a conclusion under applicable criteria; neither replaces the other.

“A technical explanation describes how a system produced or contributed to an output.A legal or academic reason explains why the authorised body reached a conclusion under applicable criteria.Neither replaces the other.”

Probably not a substantial claim: may not pass S4 (based).

Public universities are strong candidates for the category of bodies governed by public law.

“Public universities are strong candidates for the category of bodies governed by public law.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Where AI evaluates learning, recommends progression, or monitors staff performance, it can reshape substantive academic judgement.

“Where AI evaluates learning, recommends progression, or monitors staff performance, it can reshape substantive academic judgement.”

Probably not a substantial claim: may not pass S2 (checkable).

The transition from research to operational deployment must be a formal governance event, not an informal software update.

“The transition from research to operational deployment must therefore be a formal governance event, not an informal software update.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Meaningful involvement requires a reviewer who considers the relevant data and recommendation, understands the system's limits, can obtain additional evidence, has authority to change the outcome, and is not structurally induced to rubber-stamp it.

“Meaningful involvement requires a reviewer who considers the relevant data and recommendation, understands the system's limits, can obtain additional evidence, has authority to change the outcome, and is not structurally induced to rubber-stamp it.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Human oversight cannot mean that an educator is permitted to correct an output but penalised for doing so, or that a committee receives a ranked list without access to excluded files.

“Human oversight cannot mean that an educator is permitted to correct an output but penalised for doing so, or that a committee receives a ranked list without access to excluded files.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

A complex mathematical formula or exhaustive technical description is not, by itself, a meaningful explanation.

“A complex mathematical formula or exhaustive technical description is not, by itself, a meaningful explanation.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Blind or staged comparison may help test anchoring during a pilot.

“Blind or staged comparison may help test anchoring during a pilot.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The SCHUFA and Dun & Bradstreet Austria judgments support a practical explanation package.

“The SCHUFA and Dun & Bradstreet Austria judgments support a practical explanation package.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

A very low override rate can indicate either a reliable system or automation bias or lack of authority.

“A very low rate can indicate a reliable system, but it can also indicate automation bias or lack of authority.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The AI Act adds a context-specific right in Article 86.

“The AI Act adds a context-specific right in Article 86.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Intention inference requires a separate analysis because the definition in Article 3(39) is broader than the operative wording of Article 5(1)(f).

“Emotion inference from biometric data in educational institutions is generally prohibited; intention inference requires a separate analysis because the definition in Article 3(39) is broader than the operative wording of Article 5(1)(f).”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Earlier identification of students who may need support can be beneficial, but not if the model labels individuals without a support capacity, uses sensitive proxies, or creates a hidden disciplinary record.

“Earlier identification of students who may need support can be beneficial, but not if the model labels individuals without a support capacity, uses sensitive proxies, or creates a hidden disciplinary record.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

For a use already listed in Annex III, material influence and profiling determine whether Article 6(3) can exclude it from the high-risk category.

“For a use already listed in Annex III, material influence and profiling determine whether Article 6(3) can exclude it from the high-risk category.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

An institution that waits until December 2027 to discover its systems, contracts, data flows, and decision owners will have little ability to redesign a flawed workflow.

“An institution that waits until December 2027 to discover its systems, contracts, data flows, and decision owners will have little ability to redesign a flawed workflow.”

Classified as probably not a substantial claim.

The AI Act does not answer whether an HEI should automate a task simply because a compliant product exists.

“Nor does it answer whether an HEI should automate a task simply because a compliant product exists.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A proportionate implementation programme can use the runway in four phases.

“A proportionate implementation programme can use the runway in four phases.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The HEAI-AF can be implemented within existing records-management, procurement, data-protection, and quality-assurance systems; it need not be a separate software platform.

“It can be implemented within existing records-management, procurement, dataprotection, and quality-assurance systems; it need not be a separate software platform.”

May combine two claims.

Polish public and non-public universities operate under a common higher-education framework but do not have identical legal status in every context.

“Polish public and non-public universities operate under a common higher-education framework but do not have identical legal status in every context.”

Probably not a substantial claim: may not pass S2 (checkable).

Academic freedom, education, dignity, disability, equality, and effective remedy cannot be reduced to data-security controls.

“Academic freedom, education, dignity, disability, equality, and effective remedy cannot be reduced to data-security controls.”

May combine two claims.

Even before the Annex III regime applies, GDPR information and access, administrative reason-giving, equality law, contract or employment duties, and effective-remedy guarantees may independently require explanation and contestability.

“Even before the Annex III regime applies, GDPR information and access, administrative reason-giving, equality law, contract or employment duties, and effective-remedy guarantees may independently require explanation and contestability.”

May combine two claims.

Student research shows that acceptability varies substantially by data type, so generic notice or consent cannot replace contextual necessity and proportionality analysis.

“Student research also shows that acceptability varies substantially by data type, so generic notice or consent cannot replace contextual necessity and proportionality analysis (Ifenthaler & Schumacher, 2016).”

• cites Ifenthaler & Schumacher, 2016 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

The reviewer needs competence to understand the administrative matter and the model's limitations; authority to reject or change the output; information sufficient to inspect the relevant evidence; time to make a genuine assessment; independence from performance targets that reward automatic acceptance; and traceability showing what the reviewer considered and why the final decision was reached.

“The reviewer needs competence to understand the administrative matter and the model's limitations; authority to reject or change the output; information sufficient to inspect the relevant evidence; time to make a genuine assessment; independence from performance targets that reward automatic acceptance; and traceability showing what the reviewer considered and why the final decision was reached.”

Probably not a substantial claim: may not pass S2 (checkable).

The distinction between model explanation and decision reasons is central.

“The distinction between model explanation and decision reasons is central.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The Act was published on 27 July 2026 and generally entered into force on 11 August 2026, with specified provisions subject to separate dates.

“The Act was published on 27 July 2026 and generally entered into force on 11 August 2026, with specified provisions subject to separate dates: Article 125(4) entered into force on 28 July 2026”

Classified as probably not a substantial claim. May combine two claims.

The GDPR, equality law, contract and employment duties, higher-education legislation, Article 22 where triggered, and effective-remedy requirements remain relevant.

“The GDPR, equality law, contract and employment duties, highereducation legislation, Article 22 where triggered, and effective-remedy requirements remain relevant.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Principles become governable only when evidence survives staff turnover, vendor updates, complaints, and audit.

“Principles become governable only when evidence survives staff turnover, vendor updates, complaints, and audit.”

May combine two claims.

An HEI that wishes to test an admissions, assessment, or proctoring tool for disparate effect on protected groups has a clearer lawful route to the group-level validation recommended throughout this article, but the route is conditional, cumulative, and must be documented rather than assumed.

“An HEI that wishes to test an admissions, assessment, or proctoring tool for disparate effect on protected groups therefore has a clearer lawful route to the group-level validation recommended throughout this article, but the route is conditional, cumulative, and must be documented rather than assumed.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Article 1(6) of Regulation (EU) 2026/1744 inserts a new Article 4a into the AI Act governing the processing of special categories of personal data for bias detection and correction, and Article 1(9)(b) simultaneously deletes the former Article 10(5).

“Article 1(6) of Regulation (EU) 2026/1744 inserts a new Article 4a into the AI Act governing the processing of special categories of personal data for bias detection and correction, and Article 1(9)(b) simultaneously deletes the former Article 10(5).”

Classified as probably not a substantial claim.

Admission, enrolment, or programme-assignment ranking in Polish higher-education administration is a potential Annex III point 3(a) high-risk use under the AI Act.

“| Admission, | | Potential Annex III | Does the system | Equality; education; | Apply | high-risk |”

• Also stated in: § 6. Limitations and Future Research · plain

  “| enrolment, | or | point 3(a) high-risk | determine | access, | foreign-qualification | readiness; |”

• cites Regulation (EU) 2026/1744Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable).

A deferred conformity date is an implementation runway, not permission to run unclassified experiments on students or employees.

“A deferred conformity date is therefore an implementation runway, not permission to run unclassified experiments on students or employees.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Article 70 protects the right to education, while Article 73 protects freedom of scientific research, publication, teaching, and access to culture.

“Article 70 protects the right to education, while Article 73 protects freedom of scientific research, publication, teaching, and access to culture.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Admissions, student benefits, progression, expulsion, employment, and access arrangements can affect the constitutional rights to equality, privacy, informational self-determination, education, and freedom of research and teaching.

“Admissions, student benefits, progression, expulsion, employment, and access arrangements can affect the constitutional rights to equality, privacy, informational self-determination, education, and freedom of research and teaching.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The President of the Personal Data Protection Office remains the supervisory authority for GDPR compliance.

“The President of the Personal Data Protection Office remains the supervisory authority for GDPR compliance.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

A false-positive rate in remote proctoring is not merely a product metric if the result triggers a disciplinary allegation.

“A false-positive rate in remote proctoring is not merely a product metric if the result triggers a disciplinary allegation.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Articles 47 and 51 protect private life and informational self-determination.

“Articles 47 and 51 protect private life and informational self-determination.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Local repurposing can invalidate that conclusion and can change operator roles.

“Local repurposing can invalidate that conclusion and can change operator roles.”

Probably not a substantial claim: may not pass S4 (based).

A university cannot outsource legal competence or the duty to give reasons merely by procuring a predictive service.

“A university cannot outsource legal competence or the duty to give reasons merely by procuring a predictive service.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Article 32 guarantees equality before the law and prohibits discrimination.

“Article 32 guarantees equality before the law and prohibits discrimination.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

A retention model may appear supportive but become coercive if it labels students as risks, discloses sensitive inferences, or diverts human support away from those whom the model fails to identify.

“A retention model may appear supportive but become coercive if it labels students as risks, discloses sensitive inferences, or diverts human support away from those whom the model fails to identify.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Universities are autonomous institutions, but autonomy does not remove the need for judicially effective reasons.

“Universities are autonomous institutions, but autonomy does not remove publicthe need for judicially effective reasons.”

Classified as probably not a substantial claim.

Annex III identifies several education and employment uses as high-risk, subject to the conditions and exceptions in Article 6.

“Annex III identifies several education and employment uses as high-risk, subject to the conditions and exceptions in Article 6.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

File completeness, duplicate detection, or document routing, if connected to an Annex III intended purpose, may satisfy an Article 6(3) narrow procedural or preparatory condition; otherwise Article 6(3) is irrelevant.

“| File completeness, | If connected to an | Can the output | Hidden exclusion; | Obtain | provider |”

• Also stated in: § 6. Limitations and Future Research · plain

  “| duplicate detection, | Annex III intended | reject, | | delay, | corrupted | classification memo; |”

• Also stated in: § 6. Limitations and Future Research · hedged

  “| or document routing | purpose, may satisfy | prioritise, | or | documents; unequal | test edge cases; |”

• cites Regulation (EU) 2026/1744Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim. May combine two claims.

These statutes do not apply identically to every matter or every public and nonpublic HEI; the institution must identify the competent decision-maker, applicable procedure, evidential standard, accessibility duty, and route of review for the specific deployment.

“These statutes do not apply identically to every matter or every public and nonpublic HEI; the institution must identify the competent decision-maker, applicable procedure, evidential standard, accessibility duty, and route of review for the specific deployment.”

Probably not a substantial claim: may not pass S3 (load-bearing).

This source hierarchy prevented a synthetic overview from being treated as authority for requirements, such as a universal six-week Article 27 consultation or publication duty, that the enacted provision does not contain.

“This source hierarchy also prevented a synthetic overview from being treated as authority for requirements, such as a universal six-week Article 27 consultation or publication duty, that the enacted provision does not contain.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Polish higher-education and administrative law determine who has competence to decide, when an administrative decision is required, what procedural guarantees apply, and which constitutional interests must be protected.

“Polish higher-education and administrative law determine who has competence to decide, when an administrative decision is required, what procedural guarantees apply, and which constitutional interests must be protected.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Article 125(4), in force from 28 July 2026, obliges the minister responsible for computerisation to take the steps necessary to establish the relevant organisational unit within the ministry and to commence the operation of that unit, of the Commission, and of its Chairperson within two months of the Act's entry into force.

“Secondly, Article 125(4), in force from 28 July 2026, obliges the minister responsible for computerisation to take the steps necessary to establish the relevant organisational unit within the ministry and to commence the operation of that unit, of the Commission, and of its Chairperson within two months of the Act's entry into force.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

An advisory Social Council for Artificial Intelligence of nine to fifteen members operates alongside the Commission under Article 45.

“An advisory Social Council for Artificial Intelligence of nine to fifteen members operates alongside the Commission under Article 45.”

Probably not a substantial claim: may not pass S4 (based).

FRIA and DPIA protect overlapping interests through different legal tests.

“FRIA and DPIA protect overlapping interests through different legal tests.”

May combine two claims.

General FAQ chatbots and translation of published information in Polish higher-education administration are normally outside Annex III, though Article 50, AI literacy and the GDPR may apply.

“| General | FAQ | Normally | outside | Is the tool only | Incorrect | or | Limit scope; date |”

• Also stated in: § 6. Limitations and Future Research · plain

  “| chatbot | | or | Annex III; Article 50, | communicating | outdated | law; | and source content; |”

• Also stated in: § 6. Limitations and Future Research · hedged

  “| published | | GDPR may apply | information, or is it | disclosure | of | interaction | where |”

• cites Regulation (EU) 2026/1744Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

Recognition, placement, or determination of educational level in Polish higher-education administration is a potential Annex III point 3(c) high-risk use under the AI Act.

“| Recognition, | | Potential Annex III | Does the system | Devaluation | of | Preserve statutory |”

• Also stated in: § 6. Limitations and Future Research · plain

  “| placement, | or | point 3(c) high-risk | assess | | the | foreign or non- | recognition route; |”

• cites Regulation (EU) 2026/1744Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The AI Act's use of 'high-risk' and the GDPR's use of 'high risk' are not the same classification.

“The AI Act's use of”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Depending on the workflow, the legally relevant route may also involve administrative review under the Code of Administrative Procedure, labour supervision and courts under the Labour Code, equality mechanisms under the Equal Treatment Act, procurement review under the Public Procurement Law, accessibility complaints, the Commissioner for Human Rights, or the ministry responsible for higher

“Depending on the workflow, the legally relevant route may also involve administrative review under the Code of Administrative Procedure, labour supervision and courts under the Labour Code, equality mechanisms under the Equal Treatment Act, procurement review under the Public Procurement Law, accessibility complaints, the Commissioner for Human Rights, or the ministry responsible for higher”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Relevant Charter rights include human dignity (Article 1), private life and data protection (Articles 7 and 8), freedom of expression and information (Article 11), freedom of the arts and sciences and academic freedom (Article 13), education (Article 14), non-discrimination (Article 21), rights of persons with disabilities (Article 26), and an effective remedy and fair trial (Article 47).

“Relevant Charter rights include human dignity (Article 1), private life and data protection (Articles 7 and 8), freedom of expression and information (Article 11), freedom of the arts and sciences and academic freedom (Article 13), education (Article 14), non-discrimination (Article 21), rights of persons with disabilities (Article 26), and an effective remedy and fair trial (Article 47).”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The GDPR simultaneously regulates personal-data processing and certain solely automated decisions.

“The General Data Protection Regulation (GDPR) simultaneously regulates personal-data processing and certain solely automated decisions.”

May combine two claims.

Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on 27 July 2026 and moved application of the Annex III high-risk provisions to 2 December 2027.

“Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on 27 July 2026 and moved application of the Annex III high-risk provisions to 2 December 2027 (European Parliament & Council, 2026).”

• cites European Parliament & Council, 2026Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

An HEI project can face dual but complementary accountability: market-surveillance questions under the AI Act and personal-data questions under the GDPR.

“An HEI project can therefore face dual but complementary accountability: market-surveillance questions under the AI Act and personal-data questions under the GDPR.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Autonomy concerns the lawful exercise of academic and organisational competence; it is not an exemption from general law.

“Autonomy concerns the lawful exercise of academic and organisational competence; it is not an exemption from general law.”

Probably not a substantial claim: may not pass S4 (based).

An overall accuracy figure does not show whether a system systematically disadvantages applicants with disabilities, Polish-language learners, foreign qualifications, non-standard educational histories, or people who use assistive technology.

“An overall accuracy figure does not show whether a system systematically disadvantages applicants with disabilities, Polishlanguage learners, foreign qualifications, non-standard educational histories, or people who use assistive technology.”

Probably not a substantial claim: may not pass S2 (checkable).

The administrative context changes the meaning of technical performance.

“The administrative context changes the meaning of technical performance.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Procurement by institutions within the statutory scope is governed by the Public Procurement Law.

“Procurement by institutions within the statutory scope is governed by the Public Procurement Law (Republic of Poland, 2019a), and public HEIs must integrate duties under the Act on ensuring accessibility to persons with special needs (Republic of Poland, 2019b).”

• cites Republic of Poland, 2019aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

The Act of 20 July 2018 - Law on Higher Education and Science - recognises universities' legal personality and autonomy and situates the system within freedom of teaching, freedom of research, and respect for institutional self-government.

“The Act of 20 July 2018 -Law on Higher Education and Science -recognises universities' legal personality and autonomy and situates the system within freedom of teaching, freedom of research, and respect for institutional self-government (Republic of Poland, 2018).”

• cites Republic of Poland, 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The statute establishes the national institutional architecture, including the Commission for the Development and Safety of Artificial Intelligence (KRiBSI) as the market-surveillance authority and single point of contact, with operational provisions phased in under the Act.

“The statute establishes the national institutional architecture, including the Commission for the Development and Safety of Artificial Intelligence (Komisja Rozwoju i Bezpieczeństwa Sztucznej Inteligencji, KRiBSI) as the market-surveillance authority and single point of contact, with operational provisions phased in under the Act.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Responsible-AI education research supported the role-specific training component.

“Responsible-AI education research supported the role-specific training component (Aler Tubella et al., 2024).”

• cites Aler Tubella et al., 2024Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Consultation with student representatives, disability services, workers' representatives, educators, and equality or technical experts can materially improve the analysis, even when no universal consultation period is prescribed.

“Consultation with student representatives, disability services, workers' representatives, educators, and equality or technical experts can materially improve the analysis, even when no universal consultation period is prescribed.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Under Article 72 of the 2018 Act, refusal of admission is made by administrative decision and, where applicable, is governed by the Code of Administrative Procedure.

“Under Article 72 of the 2018 Act, refusal of admission is made by administrative decision and, where applicable, is governed by the Code of Administrative Procedure (Republic of Poland, 1960).”

• cites Republic of Poland, 1960Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

The Polish Constitution provides the national rights framework.

“The Polish Constitution provides the national rights framework (Republic of Poland, 1997).”

• cites Republic of Poland, 1997Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The ministry responsible for computerisation provides administrative support under Article 43(1).

“The ministry responsible for computerisation provides administrative support under Article 43(1).”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

8-18 and Chapters 3-5, 8, and 9 enter into force on 28 October 2026.

“8-18 and Chapters 3-5, 8, and 9 enter into force on 28 October 2026.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

The December 2027 Annex III application date gives Polish HEIs time to build capability; it does not justify delaying rights.

“The December 2027 Annex III application date gives Polish HEIs time to build capability; it does not justify delaying rights,”

May combine two claims.

The Article 5(1)(f) prohibition relevant to educational emotion inference and the Article 4 AI-literacy duty have applied since 2 February 2025; specified Article 50 transparency duties apply from 2 August 2026; and the application dates of later amendments must be checked provision by provision.

“The Article 5(1)(f) prohibition relevant to educational emotion inference and the Article 4 AI-literacy duty have applied since 2 February 2025; specified Article 50 transparency duties apply from 2 August 2026; and the application dates of later amendments must be checked provision by provision.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

Formal establishment is not operational capacity, and the phased dates in Articles 125 and 127 should be read accordingly.

“Thirdly, formal establishment is not operational capacity, and the phased dates in Articles 125 and 127 should be read accordingly.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Article 41 of the Charter is addressed to Union institutions and cannot simply be invoked as a universal cause of action against every university.

“Article 41 of the Charter is addressed to Union institutions and cannot simply be invoked as a universal cause of action against every university; however, legality, impartiality, reasongiving, and the opportunity to be heard also arise through national administrative law, general principles, and Article 47 where Union rights are engaged.”

May combine two claims.

The GDPR, Polish constitutional guarantees, equality norms, higher-education legislation, labour law, procurement duties, and procedural safeguards continue to operate.

“The GDPR, Polish constitutional guarantees, equality norms, higher-education legislation, labour law, procurement duties, and procedural safeguards continue to operate.”

May combine two claims.

Legality, impartiality, reason-giving, and the opportunity to be heard also arise through national administrative law, general principles, and Article 47 where Union rights are engaged.

“Article 41 of the Charter is addressed to Union institutions and cannot simply be invoked as a universal cause of action against every university; however, legality, impartiality, reasongiving, and the opportunity to be heard also arise through national administrative law, general principles, and Article 47 where Union rights are engaged.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Profiling does not create an Annex III listed purpose, but it prevents an Annex III system from relying on the Article 6(3) exclusion.

“Profiling does not create an Annex III listed purpose, but it prevents an Annex III system from relying on the Article 6(3) exclusion.”

Classified as probably not a substantial claim.

Evaluation of learning outcomes or progression recommendation in Polish higher-education administration is a potential Annex III point 3(b) high-risk use under the AI Act.

“| Evaluation | of | Potential Annex III | Does the output | Academic freedom; | Define assessment |”

• cites Regulation (EU) 2026/1744Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Employment uses must also be assessed under the Labour Code.

“Employment uses must also be assessed under the Labour Code (Republic of Poland, Poland, 2010).”

• cites Republic of Poland, Poland, 2010Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Academic freedom is relevant not only when AI generates teaching content but also when an optimisation system changes educators' discretion, student-teacher relations, or the evidential status of academic judgement.

“Academic freedom is relevant not only when AI generates teaching content but also when an optimisation system changes educators' discretion, student-teacher relations, or the evidential status of academic judgement.”