Legal Setting and Supervision

This chapter sets out the legal and supervisory setting in which Nordvik Region works: the jurisdiction, the instruments that apply, the role the Region has under each, the supervisors it answers to and the way data crosses its boundaries. It draws on every volume that names a legal fact about the Region, above all AI Governance, which fixes the profile, and on the later volumes that apply it. The statements are attributed to the instrument and the article as the volumes attribute them, and they are framed as organisational and architectural implications, not as legal advice. Article numbers that the volumes cite and that this chapter relies on for dates were checked against the consolidated text of the AI Act, which includes Regulation (EU) 2026/1744. The facts about the Region are fictional, and the legal instruments are real.

One jurisdiction, public law

The Region works under one jurisdiction, and the series treats this as one of its advantages: one legal regime, a clear public mandate and a long tradition of openness. The profile gives the legal setting as five strands (2, Nordqvist, 2026).

Strand What the profile says
The AI Act Deployer of high-risk systems (Annex III point 5(a), access to essential public assistance benefits and services, as Landscape words the point), with the deployer duties of Article 26, the fundamental rights impact assessment of Article 27 and registration as a public-authority deployer under Article 49. Article 4 AI literacy and Article 50 transparency for its assistants
The GDPR Article 6(1)(e) as the usual legal basis; special-category data under Article 9 in health and social care; the national supervisory authority
National administrative law The duty to give reasons, the right to be heard, case handling by an accountable official; public-access and archives law; public procurement law (Directive 2014/24/EU as transposed)
NIS 2 As transposed nationally, "to the extent regional government entities fall within scope"
Labour law One national co-determination regime; employee representation (Chapter 3)

The profile does not name the state, and AI Environment writes Sweden. Later volumes use Swedish sources as the example for Nordvik Region: the Co-determination Act (sections 11, 12, 19, 38 and 64) in AI Boss, AI Operations and AI Developer, the Local Government Act (chapter 6 section 37 on an automated decision function, and chapter 12 section 1 on the auditors) in AI Leader, and a Cybersecurity Act with the term "essential operator" in AI Operations and AI Leader. AI Operations states in terms that the book sets the Region in Sweden (7, Nordqvist, 2026) (10, Nordqvist, 2026) (11, Nordqvist, 2026). The dossier keeps the profile's wording, "a Nordic member state", and records the Swedish examples as the volumes use them. This is an open thread for Chapter 27.

The legal setting of Nordvik Region: the instruments, the roles the Region has under them and the supervisors to which each leads.

Figure 4.1: Instruments, roles and supervisors. The Region is deployer under the AI Act and provider only of the explanation service it built.

The AI Act and the Region's roles

Deployer of a high-risk system

The Region is a deployer of the benefits-triage system, which ranks applications for financial assistance by urgency and is high-risk under Annex III point 5(a). AI Governance draws the consequences. Article 26 gives the deployer duties: to assign human oversight to natural persons with the necessary competence, training and authority, to monitor the system, to keep logs for at least six months and to inform workers' representatives before a high-risk system is used at the workplace. Article 27 requires a fundamental rights impact assessment, and Article 49 requires registration. The Region recorded the fundamental rights impact assessment as FRIA-2026-002, completed on 2026-06-20 and notified to the market surveillance authority on 2026-07-02. It registered the system in the EU database as EUDB-DEP-2026-004213, and published a public register entry NR-ALG-2026-003 for the living-support triage with a reasons log RL-2026-0481733 (2, Nordqvist, 2026). The oversight design is a decision record, AIDR-0011, and it sets the figures that make oversight a mechanism: a ranking limit of 40 an hour, a minimum of 45 seconds a case, a blind sample of 5 per cent and override alarms between 2 and 30 per cent.

The deployer duties apply from 2 December 2027 for Annex III systems, under Article 113 as amended by Regulation (EU) 2026/1744, and that date appears in several volumes as the review trigger for the Region's plans (2, Nordqvist, 2026) (3, Nordqvist, 2026) (7, Nordqvist, 2026). For systems already in use, Article 111(2) as consolidated requires providers and deployers of high-risk systems intended for use by public authorities to take the necessary steps by 2 August 2030, and AI Leader uses that date (11, Nordqvist, 2026). The Region's first plan in AI Governance closes its Article 26 evidence gap before 2 December 2027. Article 113, third paragraph, point (c), as read in the consolidated text, names Chapter III, Sections 1, 2 and 3 (which hold Articles 26 and 27) for the date of 2 December 2027; Article 49 sits in Section 5 and is not named there. Landscape therefore no longer groups the registration of Article 49 with the deferred duties, and the dossier records the point without drawing a date from it, since no volume gives the Region a separate registration date under the amended text (15, Nordqvist, 2026).

Provider of one service

The Region is not only a deployer. AI Governance records that it built the benefits-explanations service around the claims ranker in house and is provider and deployer of that service only (2, Nordqvist, 2026). AI Platform and Security confirms the pattern, and its staff assistant is entered as compoundSystem provider, generalPurposeModel deployer, since the Region built the system around a general-purpose model (3, Nordqvist, 2026). AI Boss adds a further case. The home-care dashboards, built in house, may or may not fall under Annex III point 4(b), which concerns systems used to allocate tasks or monitor and evaluate workers. The Region decided in June 2027 (AIDR-0609, illustrative) to treat them as high-risk by precaution and so to carry provider duties, which makes a small digitalisation unit a provider (7, Nordqvist, 2026).

Article What the Region does with it Volume
Article 4, AI literacy A programme that reaches about 6,000 employees in one language; 45 minutes before access and an annual refresher; a builder learning plan BLP-NR-2031-01 AI Governance; Vibe to Dev
Article 5(1)(c) A prohibited-practice check PPC-NR-01 stopped a proposed travel-card feature for the claims ranker AI Threat
Article 26(2) Named oversight persons: caseworker team leaders and a senior caseworker, six of six trained AI Governance
Article 26(5) The provider is informed when an incident touches a deployed system; done 45 minutes after the declaration in INC-NR-2028-0163 (illustrative) AI Red Team; AI Operations
Article 26(6) A log floor of six months; the Region's tender annex requires 12 months (SR-02) and AIDR-0513 at least 183 days The Transformed Security Architect; AI Governance
Article 26(7) Information to workers' representatives before a workplace system; the Region applied the high-risk information procedure voluntarily AI Governance; AI Boss
Article 43(4) Removing a feature lies outside the model's change envelope and needs a new record, a risk assessment and an updated fundamental rights assessment AI Platform and Security
Article 50 Disclosure by the citizen-service assistant; an open obligation OBL-NR-08 on the disclosure of deepfake content under Article 50(4) AI Platform and Security; AI Purple
Article 86 Cited with Articles 26, 27 and 49 for the benefits-triage system AI Threat

The Article 26(5) interval is derived from the incident timeline of AI Operations, where the incident was declared at 09:20 and the provider was informed at 10:05 (illustrative) (10, Nordqvist, 2026).

Across the usage modes

The legal consequence of a system depends on its mode, and the volumes treat each mode in its own way.

Mode Legal and contractual consequence in the volumes
Mode 0 Ambient AI No owner before 2027-06-30 (illustrative, Landscape). AI Red Team sets supplier requirements NR-SUP-SECAI-001: declare AI functions, the model provider and version, automatic actions and the effective autonomy level; 30 days' notice of change; an unnotified change is a breach; six months for exit
Mode 1 Conversational Article 4 literacy; data standard STD-AI-002 v1.3 (public data in any approved tool; internal only in the Region's tenant; confidential only in the staff assistant; sensitive in no system without a data protection impact assessment); consumer endpoints blocked
Mode 2 Embedded A pilot with a model chosen by the suite supplier under a non-EU jurisdiction, classed accept_and_monitor in AI Threat; the profile places it in Mode 2
Mode 3 Agentic work tools Only public and internal material on an approved cloud model; anything above internal on a pinned self-hosted model
Mode 4 Applications Deployer duties under Articles 26, 27 and 49 where high-risk; transparency under Article 50 for assistants; a data protection impact assessment; a public register entry
Mode 5 Not permitted by policy

(4, Nordqvist, 2026) (5, Nordqvist, 2026) (2, Nordqvist, 2026) (3, Nordqvist, 2026) (15, Nordqvist, 2026).

Data protection

The Region bases most processing on Article 6(1)(e) of the GDPR, a task carried out in the public interest, and handles special categories of data under Article 9 in health and social care. This has architectural consequences that the volumes spell out. The case-note summariser for 180 social-services caseworkers runs on premises with all egress denied. The security assistant is self-hosted because alarm data contains personal and health data. Findings and threat models are kept in the Region's own tenant, and a lawful mandate LM-2026-009 (2026-09-22) covers the intake use case UC-2026-017 (2, Nordqvist, 2026) (3, Nordqvist, 2026) (9, Nordqvist, 2026).

The data protection impact assessments are DPIA-2026-007 for the benefits-triage system and DPIA-2026-031 for rostering. The data protection officer is a reviewer or adviser of nearly every decision record and receives changes to the staff assistant's prompt. In AI Boss, the negotiated rule that open shifts are offered in a published rotating order, and not ranked, was recorded as AIDR-0626 under Article 22 of the GDPR, which restricts decisions based solely on automated processing (7, Nordqvist, 2026). Personal data incidents are reported to the data protection authority, and in INC-NR-2028-0163 the GDPR clock was closed at 15:30 with no breach (illustrative) (4, Nordqvist, 2026) (10, Nordqvist, 2026).

NIS 2, DORA and other instruments

NIS 2. The profile says that the position "depends on national transposition". AI Red Team maps its CSF profile NR-CSF-2027 to NIS 2 and its national transposition, and reports significant incidents to the national computer security incident response team (CSIRT) and takes part in a regional-government information-sharing group (4, Nordqvist, 2026). The managed security provider is covered by Implementing Regulation 2024/2690, which the Region uses as a checklist for what it requires of the supplier. The Transformed Security Architect trains the executive committee under Article 20 of NIS 2 (AIDR-0513), and AI Blue is unsure whether the executive committee counts as a management body, which it says was confirmed with counsel (6, Nordqvist, 2026) (8, Nordqvist, 2026). AI Operations moves from the profile's conditional position to a decision. The Region plans as if its Tier 1 services were in scope, and AIDR-0923 (2028-11-21, illustrative) applies the Cybersecurity Act to all Tier 1 services as an "essential operator". In INC-NR-2028-0163 the NIS 2 clock was closed at 14:00 with the incident assessed as not significant (10, Nordqvist, 2026). The conditional and the decided positions are an open thread.

DORA. The Region is not covered by Regulation (EU) 2022/2554. AI Operations says the Digital Operational Resilience Act does not apply to it, and AI Developer treats Article 9(4)(e) and Delegated Regulation (EU) 2024/1774 Article 17 as a defensible standard that the Region may follow without being bound (10, Nordqvist, 2026) (12, Nordqvist, 2026).

The Cyber Resilience Act and product liability. AI Developer records that the Region is not a manufacturer under Regulation (EU) 2024/2847 and may be an "open-source steward" for the citizen-service components it has published under an open licence since 2027. Reporting obligations apply from 2026-09-11 and manufacturer obligations from 2027-12-11, and the revised Product Liability Directive applies to products placed after 2026-12-09 (12, Nordqvist, 2026).

Data-centre energy reporting. AI Environment gives the Region a further legal footing, all of it illustrative in its dates: nordvik-dc has an installed IT power of about 640 kW, above the 500 kW threshold of Article 12 of the Energy Efficiency Directive, so the Region reports as an operator under the Swedish Lag (2025:570) (LGL-NR-0001); the Region's hosting provider reports under the same Article, and the Region holds its Annex VII figures for reporting year 2031 (received 2032-05-15). LGL-NR-0002 records Article 53(1)(a) of the AI Act on the technical documentation of a general-purpose AI model provider, and LGL-NR-0003 the Commission's rating act C(2026) 3472 final (adopted, not in force), all read as of 2026-11-16. The Region is a public authority and not a reporting entity under the Corporate Sustainability Reporting Directive; it reports energy and climate in its annual report, and its register records on AI uses, with footprint and benefit, are requestable under the public-access rules unless classified. These are architectural implications taken from the volume; the dossier verified Article 53 of the AI Act against the consolidated text, and the Energy Efficiency Directive and the Swedish law are not among the instruments the fetch tool serves, so they are taken from the volume (14, Nordqvist, 2026).

Critical entities and energy. AI Operations records that the Region was not identified as a critical entity at the start of the book (illustrative), with exposure through its dependence on the university hospital of the neighbouring region, and that its two data-centre halls, at an assumed 640 kW, exceed the 500 kW threshold for reporting under the Energy Efficiency Directive (10, Nordqvist, 2026).

Platform work. The transposition deadline of the Platform Work Directive passed on 2026-12-02. The Region runs no platform, and AI Boss reads its chapter III as a benchmark for algorithmic management (7, Nordqvist, 2026).

Public procurement, public access and administrative law

Procurement. The Region buys most of its AI, and so Directive 2014/24/EU is a design instrument. The Transformed Security Architect treats the contract as an architectural instrument under Articles 42 and 70, and puts the stop path, the default-deny egress and the credentials of under 60 minutes into the tender annex TA-NR-S4 and the requirement set SECREQ-NR-2026-117 (6, Nordqvist, 2026). AI Boss writes the same lesson into NR-PRQ-MGT-01: restrictions as feasibility conditions, a technical stop within one hour with a manual fallback, a log kept 24 months, outcome exports by contract type, sex and age band, an award criterion and a change clause, under Articles 42, 67, 70 and 72 (7, Nordqvist, 2026). A tender above EUR 2 million needs a decision of the executive committee, and procurement lead times run from 9 to 14 months (11, Nordqvist, 2026) (10, Nordqvist, 2026). Chapter 6 treats the suppliers.

Public access and archives. The public-access principle and the archives law shape every artefact. Decision records and management decision records are written as if they could be requested. The retention schedule names AI-generated notes, drafts and transcripts. The detailed registers of security AI, threat models, exercise findings and decoys are classified, and the threat register is published with a classified annex for two of its entries (2, Nordqvist, 2026) (4, Nordqvist, 2026) (5, Nordqvist, 2026). Whether prompts are records is open for the archivist and the counsel in AI Developer, and Vibe to Dev keeps prompts in the assistant's log for 30 days under an architecture decision record agreed with the archivist (12, Nordqvist, 2026) (13, Nordqvist, 2026).

Administrative law. A named caseworker must decide each case and give reasons. AI Leader adds that the delegation order mentions no AI, that its proposed change records that no automated decision function exists under the Local Government Act, and that the ranking of the triage system must never be the ground for a decision or a reason (AIDR-1018, 2030-02, illustrative) (11, Nordqvist, 2026).

Supervisors and who reports to whom

Supervisor or body What the Region reports or submits Volume
Market surveillance authority FRIA-2026-002 notified 2026-07-02; an earlier summary for the living-support triage notified 2026-02-18 AI Governance
EU database Registration of the high-risk deployer; Article 49(3) registration expected for rostering AI Governance; AI Boss
Data protection authority Personal data incidents; the GDPR clock in incidents AI Red Team; AI Operations
National CSIRT and the NIS 2 authority Significant incidents under the national transposition AI Red Team; AI Operations
Regional-government information-sharing group Participation, traffic light protocol green (AIDR-0831, 2028-04-11) AI Red Team; AI Purple
National channel for data-centre operators and the European data-centre database Annual Annex VII report of nordvik-dc under Article 12 of the Energy Efficiency Directive and Lag (2025:570), signed by the property and energy manager (illustrative) AI Environment
Audit committee, elected auditors, internal audit Threat register at each half-year; assurance questions AI Threat; AI Purple; AI Leader
Public algorithm register NR-ALG-2026-003 (living-support triage), NR-ALG-2026-007 (staff assistant) and the replacement NR-ALG-2026-009 for the retired travel-reimbursement scoring NR-ALG-2022-011 AI Governance

(2, Nordqvist, 2026) (4, Nordqvist, 2026) (9, Nordqvist, 2026).

Transfers, residency and sovereignty

The Region's rule is that data stays inside the European Economic Area, and the volumes apply it by class. The staff assistant runs on an open-weight model at a European cloud provider under contract NR-2026-031, with processing in the EEA only and keys held by the customer. Security data never leaves the EU or the EEA, which the charter records as securityDataLeavesEuEea: false. The managed security provider must process and store logs inside the EU or the EEA under framework agreement NR-2025-044. The origin and hosting policy adopted in 2026 requires suppliers to name all sub-processors, to report government requests where law allows and to give 180 days' notice before withdrawing a model version. AI Platform and Security adds a sovereignty profile for a managed-inference tender: EEA processing (SOV-01), no training on inputs (SOV-04), open weights with export at exit (SOV-05) and exit assistance of at least 12 months (SOV-08) (2, Nordqvist, 2026) (3, Nordqvist, 2026) (4, Nordqvist, 2026). The decision on the staff assistant's model origin, AIDR-0211, chose an open-weight model at a European provider for data protection and exit, with a proprietary hyperscaler in an EU region kept as a fallback for non-personal tasks. Legal counsel recorded a note on it as LC-2026-044. Chapter 6 gives the supplier and sovereignty picture in full.

Some of the Region's AI sits outside the EU's jurisdiction, and the volumes do not hide it. The recruitment screening pilot came from a supplier outside the EU, the embedded workplace assistant uses a model chosen by a suite supplier outside the EU, and licensed software carries supplier-enabled AI from outside the EU. Landscape draws four of its units outside the EU in its supply view (5, Nordqvist, 2026) (15, Nordqvist, 2026).

Open threads

  • Fundamental rights assessments. FRIA-2026-002 (complete 2026-06-20, notified 2026-07-02) is for the benefits-triage system, and a fundamental rights summary for NR-ALG-2026-003 (version 2, complete 2026-02-11, notified 2026-02-18) is for the living-support triage, which the profile says the same claims ranker orders. AI Governance also describes a separate "lsb-triage" system in the later chapter. It is unclear whether the two are one system (2, Nordqvist, 2026).
  • EU database identifiers. EUDB-DEP-2026-004213 and EUDB-2026-NR-00412 for the benefits-triage system, in AI Governance Chapters 9 and 17.
  • Provider or deployer. The use-case register NR-UC-020 enters the staff assistant with ai_act_role: provider, and UCD-2026-022 enters it as deployer, in AI Governance; AI Platform and Security has the compound role above.
  • Waiver length and retention. A waiver of at most 90 days against 180 days for high risk, STD-AI-002 at 183 days against the text "six months", and log retention of 12 months (RET-04), 12 months (SR-02), 120 months (evidence) and 13 months (archive) in different volumes. Chapter 23 collects the figures.
  • NIS 2 position and the state. Conditional on transposition in the profile and decided as "essential operator" in AIDR-0923; the earlier volumes leave the state unnamed while using Swedish sources, and AI Environment names Sweden.
  • Installed power of nordvik-dc. AI Operations assumes 640 kW and says the halls exceed the 500 kW threshold; AI Environment gives about 640 kW as new, illustrative profile data and a power usage effectiveness of about 1.55. The values agree, and the status moves from an assumption to a profile figure in a later volume (10, Nordqvist, 2026) (14, Nordqvist, 2026).
  • Notice periods. 180 days for a model version withdrawn (AI Governance), 60 days for a model change at Eastby, 20 working days in the tender annex of The Transformed Security Architect and in annex MS-2, and 30 days in NR-SUP-SECAI-001. They are for different suppliers, and the dossier does not claim a conflict, but a reader should not take one for another.

Summary

Nordvik Region works under one jurisdiction and one body of public law. It is a deployer of one high-risk system, a provider of one service it built and, by precaution, a provider of dashboards that it treats as high-risk, and it carries the duties that follow from each. The AI Act deployer duties apply to Annex III systems from 2 December 2027, with 2 August 2030 as the outer date for systems of public authorities already in use. Procurement law, public access, archives and administrative law turn the contract, the record and the reasons into architectural instruments, and employee representation requires negotiation before use. DORA does not apply, NIS 2 moves from conditional to decided in the later volumes, and data stays in the European Economic Area by class and by contract, with some AI sourced from outside the EU. The dossier's own conclusion is that the Region's legal setting is simpler in jurisdiction and heavier in public-law duty than that of Hansa Mutual, that most of its obligations are met through the contract, and that the open threads are mostly differences of date, identifier and role between volumes, not differences of law.

Sources

  1. Nordqvist, G. AI Governance: Governing Artificial Intelligence in European Enterprises and Public Authorities. (2, Nordqvist, 2026)
  2. Nordqvist, G. AI Platform and Security: Building and Securing AI Platforms in European Enterprises and Public Authorities. (3, Nordqvist, 2026)
  3. Nordqvist, G. AI Red Team: AI as an Active Part of Cyber Defence in European Enterprises and Public Authorities. (4, Nordqvist, 2026)
  4. Nordqvist, G. AI Threat: Are the Dystopias Coming True? Evidence, Risk and Response for European Enterprises and Public Authorities. (5, Nordqvist, 2026)
  5. Nordqvist, G. The Transformed Security Architect: Role, Craft and Career in the Age of AI in European Enterprises and Public Authorities. (6, Nordqvist, 2026)
  6. Nordqvist, G. AI Boss: When Artificial Intelligence Manages People in European Enterprises and Public Authorities. (7, Nordqvist, 2026)
  7. Nordqvist, G. AI Blue: How Artificial Intelligence Transforms the Blue Team in European Enterprises and Public Authorities. (8, Nordqvist, 2026)
  8. Nordqvist, G. AI Purple: How Red and Blue Learn Together When AI Joins the Loop in European Enterprises and Public Authorities. (9, Nordqvist, 2026)
  9. Nordqvist, G. AI Operations: How Artificial Intelligence Transforms IT Operations in European Enterprises and Public Authorities. (10, Nordqvist, 2026)
  10. Nordqvist, G. AI Leader: How Leadership Roles Change When AI Is Used Broadly in European Enterprises and Public Authorities. (11, Nordqvist, 2026)
  11. Nordqvist, G. AI Developer: How the Developer Role Changes When AI Is Used Broadly in European Enterprises and Public Authorities. (12, Nordqvist, 2026)
  12. Nordqvist, G. Vibe to Dev: From Vibe Coding to Professional Software, Step by Step, for Builders Who Are Not Developers, in European Enterprises and Public Authorities. (13, Nordqvist, 2026)
  13. Nordqvist, G. AI Environment: Designing AI for the Greatest Benefit and the Least Environmental Harm in European Enterprises and Public Authorities. (14, Nordqvist, 2026)
  14. Nordqvist, G. Landscape: A Fact-Based Picture of Architecture and AI at Work, for Leaders, Architects and Citizens. (15, Nordqvist, 2026)