CSOAI OS · the regulator atlas

Every regulator. The tools and the clock.

The major AI and cybersecurity regimes worldwide — each with the top 7 tools you need and the next 7 dates that matter. The Council assistant gives a live read on any of them, then does the work: classify, assess, sign.

You are interacting with an AI system.

The atlas Q&A sends your input to the live Council chat endpoint (councilof.ai/api/gspc), where a model writes the answer. The Art 50(1) notice for this surface is registered here and being wired; until the component ships, this registry entry is the disclosure.

Disclosed under EU AI Act Article 50(1). Every surface and its classification

EU AI Act
European Union · European Commission · AI Office · national authorities
Phasing in

The first comprehensive, binding, risk-tiered AI law. Prohibited practices, high-risk obligations, transparency duties, and GPAI rules — phasing in through 2027.

Top 7 tools needed
  1. 1.Risk classification engine (prohibited / high-risk / limited / minimal)
  2. 2.Conformity assessment + CE-marking workflow (Art. 43)
  3. 3.Technical documentation pack (Annex IV)
  4. 4.Fundamental Rights Impact Assessment (FRIA, Art. 27)
  5. 5.Event logging & traceability (Art. 12)
  6. 6.Human-oversight controls (Art. 14)
  7. 7.Post-market monitoring + serious-incident reporting (Art. 72/73)
Next 7 dates & movements
  • 2 Dec 2026Art. 50 transparency duties apply (moved from Aug 2026 by the Digital Omnibus — verify final text)
  • Q3–Q4 2026GPAI Code of Practice adherence reviewed; AI Office guidance expands
  • 2 Dec 2027High-risk Annex III obligations apply (Digital Omnibus timeline)
  • 2026–27Harmonised CEN/CENELEC standards finalised for high-risk conformity
  • ongoingAI Office enforcement of GPAI systemic-risk models (>10^25 FLOP)
  • ongoingMember-State penalty regimes + market-surveillance authorities stand up
  • 2028+Commission review of Annex III scope + prohibited-practice list
How CSOAI covers it

Full obligation map, FRIA + conformity workflow, Art. 12 logging signed to Layer 0, and a live deadline clock. Comply once — crosswalk to NIST + ISO 42001. (Dates track the 2026 Digital Omnibus — verify against the final published text.)

Open in the Hive →Govern a case →
NIST AI RMF
United States · NIST (voluntary framework)
Voluntary

The AI Risk Management Framework 1.0 plus the Generative AI Profile — voluntary, function-based (Govern, Map, Measure, Manage). With EO 14110 revoked (Jan 2025), the operative US federal baseline is OMB M-25-21 (which rescinded and replaced M-24-10 in April 2025) + NIST AI RMF, layered with state law (Colorado, California).

Top 7 tools needed
  1. 1.Govern/Map/Measure/Manage program scaffold
  2. 2.AI system inventory + context mapping
  3. 3.Bias & fairness measurement suite
  4. 4.Model cards + data sheets
  5. 5.Adversarial testing / red-teaming
  6. 6.Incident + near-miss tracking
  7. 7.Third-party / supply-chain risk controls
Next 7 dates & movements
  • ongoingGenAI Profile (NIST-AI-600-1) adoption across agencies + vendors
  • 2026Sector crosswalks + NIST assurance guidance expand
  • in effectOMB M-25-21 governs federal agencies' AI use (CAIOs, risk practices) — rescinded and replaced M-24-10 in Apr 2025; EO 14110 revoked Jan 2025
  • ongoingFederal procurement + M-25-21 increasingly reference RMF conformance
  • 2026–27Alignment work between RMF and ISO/IEC 42001 controls
  • 1 Jan 2027State AI laws begin to bite (e.g. Colorado SB 26-189 / ADMT)
  • ongoingCISA + sector guidance for AI in critical infrastructure
How CSOAI covers it

The Govern/Map/Measure/Manage scaffold as a live program, model cards + bias audits, red-team logging — mapped to EU AI Act so one evidence set serves both.

Open in the Hive →Govern a case →
ISO/IEC 42001
Global · ISO/IEC · accredited certification bodies
In force

The first certifiable AI management system standard (2023). The 'ISO 27001 for AI' — auditable, globally recognised, and increasingly demanded in enterprise procurement.

Top 7 tools needed
  1. 1.AI Management System (AIMS) documentation
  2. 2.Statement of Applicability (Annex A controls)
  3. 3.AI risk assessment + treatment plan
  4. 4.AI system impact assessment
  5. 5.Lifecycle & data-governance controls
  6. 6.Internal audit + management review
  7. 7.Continual improvement / nonconformity log
Next 7 dates & movements
  • ongoingRising enterprise + government procurement demand for 42001 certification
  • 2026Guidance standards (42005 impact assessment, 42006 audit) mature
  • ongoing42001 ↔ EU AI Act conformity mapping tightens
  • 2026–27Accredited certification-body capacity expands globally
  • ongoingIntegration with ISO 27001 / 27701 management systems
  • 2027+First scheduled review cycle of the standard
  • ongoingSector profiles (health, finance) emerging
How CSOAI covers it

AIMS + Statement of Applicability generated from your controls, impact assessments, and audit logs — exportable and signed. One certification, mapped across regimes.

Open in the Hive →Govern a case →
GDPR
European Union · EDPB · national DPAs
In force

The data-protection backbone AI systems must respect — lawful basis, purpose limitation, automated-decision rights (Art. 22), and DPIAs where processing is high-risk.

Top 7 tools needed
  1. 1.Records of Processing Activities (RoPA)
  2. 2.Data Protection Impact Assessment (DPIA)
  3. 3.Lawful-basis + consent management
  4. 4.Art. 22 automated-decision safeguards
  5. 5.Data-subject-request (DSAR) workflow
  6. 6.Data minimisation + retention controls
  7. 7.Cross-border transfer mechanism (SCCs/adequacy)
Next 7 dates & movements
  • ongoingDPAs scrutinising training-data scraping + generative AI
  • 2026EDPB opinions on AI models & personal data mature
  • ongoingGDPR ↔ EU AI Act interplay clarified for high-risk systems
  • ongoingEnforcement on biometric + emotion-recognition data
  • 2026GDPR procedural regulation streamlining cross-border cases
  • ongoingAutomated-decision case law (Art. 22) expanding
  • ongoingRising fines for AI-driven profiling failures
How CSOAI covers it

DPIA + RoPA generation, Art. 22 safeguards, and DSAR workflows — with the EU AI Act FRIA sharing evidence so you don't assess twice.

Open in the Hive →Govern a case →
Colorado AI framework (SB 26-189 / ADMT)
United States · Colorado · Colorado Attorney General
Phasing in

Originally the first comprehensive US state AI law (SB 24-205 — a duty of reasonable care against algorithmic discrimination in consequential decisions), but repealed and replaced in May 2026 by SB 26-189, a scaled-back ADMT framework that drops the duty of care and impact-assessment mandates. Effective 1 Jan 2027. The requirements below reflect the original SB 24-205 regime.

Top 7 tools needed
  1. 1.High-risk consequential-decision inventory
  2. 2.Algorithmic-discrimination risk management program
  3. 3.Consumer notice + explanation workflow
  4. 4.Impact assessments (annual + on material change)
  5. 5.Right-to-appeal / human review path
  6. 6.AG disclosure of discovered discrimination
  7. 7.Developer↔deployer documentation exchange
Next 7 dates & movements
  • 1 Jan 2027SB 26-189 (ADMT framework) takes effect — repealed & replaced the original SB 24-205 AI Act, dropping its duty of care and impact-assessment mandates
  • 2026AG rulemaking under the scaled-back ADMT framework
  • ongoingOther US states watching Colorado's pivot away from the EU-style model
  • ongoingInteraction with NIST RMF as the reasonable-care benchmark
  • 2026–27Enforcement posture + safe-harbour clarifications
  • ongoingBusiness-community amendments debated in legislature
  • TBDPossible federal preemption discussions — status shifting
How CSOAI covers it

Consequential-decision inventory, algorithmic-discrimination testing, and consumer-notice + appeal workflows — reusing your NIST/EU evidence.

Open in the Hive →Govern a case →
NIS2 Directive
European Union · National cyber authorities · ENISA
In force

The EU's expanded cybersecurity directive — risk-management measures, incident reporting, and management accountability for essential and important entities across 18 sectors.

Top 7 tools needed
  1. 1.Asset + supply-chain risk register
  2. 2.Cyber risk-management measures (Art. 21)
  3. 3.24h/72h incident-reporting workflow
  4. 4.Business continuity + backup controls
  5. 5.Vulnerability disclosure + patch management
  6. 6.Management-body oversight + training
  7. 7.Supply-chain security assessments
Next 7 dates & movements
  • ongoingNational transposition + enforcement ramping (deadline was Oct 2024)
  • 2026Registration of essential/important entities completed across states
  • ongoingImplementing acts on technical + methodological requirements
  • ongoingFirst enforcement actions + management-liability tests
  • 2026–27Alignment with CRA + DORA for overlapping entities
  • ongoingENISA guidance + sectoral thresholds refined
  • ongoingCross-border incident coordination via CSIRTs network
How CSOAI covers it

Cyber risk register, incident-reporting clock, and supply-chain assessments — plus cyber self-scan so you can test controls, not just document them.

Open in the Hive →Govern a case →
DORA
European Union · ESAs (EBA/ESMA/EIOPA) · national regulators
In force

The Digital Operational Resilience Act — ICT risk management, incident reporting, resilience testing, and third-party (incl. cloud/AI) oversight for EU financial entities.

Top 7 tools needed
  1. 1.ICT risk-management framework
  2. 2.ICT-incident classification + reporting
  3. 3.Digital operational resilience testing (incl. TLPT)
  4. 4.ICT third-party register + concentration risk
  5. 5.Contractual clauses for critical ICT providers
  6. 6.Business-continuity + response/recovery plans
  7. 7.Board-level ICT governance
Next 7 dates & movements
  • ongoingFull application in effect (since Jan 2025) — supervisory ramp-up
  • 2026Register-of-information submissions + oversight of critical TPPs
  • ongoingThreat-led penetration testing (TLPT) cycles begin for larger entities
  • ongoingRTS/ITS technical standards bedding in
  • 2026–27Designation + oversight of critical ICT third parties
  • ongoingCross-mapping with NIS2 for dual-scope entities
  • ongoingAI-vendor dependency treated as ICT third-party risk
How CSOAI covers it

ICT risk framework, incident classification, and third-party (AI/cloud) register — resilience testing evidence signed to Layer 0.

Open in the Hive →Govern a case →
Cyber Resilience Act
European Union · European Commission · market surveillance
Phasing in

Mandatory cybersecurity requirements for products with digital elements — secure-by-design, vulnerability handling, and CE-marking for hardware + software across the EU market.

Top 7 tools needed
  1. 1.Product security risk assessment
  2. 2.Secure-by-design + secure-by-default controls
  3. 3.SBOM (software bill of materials)
  4. 4.Coordinated vulnerability disclosure process
  5. 5.Security update / patch delivery mechanism
  6. 6.Conformity assessment + CE marking
  7. 7.Actively-exploited-vuln + incident reporting
Next 7 dates & movements
  • Sept 2026Vulnerability + incident reporting obligations begin to apply
  • Dec 2027Full CRA obligations apply for products with digital elements
  • 2026Harmonised standards + guidance for essential requirements
  • ongoingNotified-body capacity for conformity assessment builds out
  • 2026–27SBOM tooling + attestation expectations mature
  • ongoingOverlap handling with AI Act for AI-enabled products
  • ongoingOpen-source steward obligations clarified
How CSOAI covers it

SBOM + secure-by-design checklist, CVD process, and conformity workflow — with cyber self-scan checking your product surface for the evidence.

Open in the Hive →Govern a case →
China AI Rules (TC260 / GenAI Measures)
China · CAC · TC260
In force

Interim Measures for Generative AI plus TC260 standards and content-labelling rules — security assessments, training-data governance, and provider registration.

Top 7 tools needed
  1. 1.Algorithm + service filing/registration
  2. 2.Security self-assessment
  3. 3.Training-data + content governance
  4. 4.AI-generated-content labelling
  5. 5.Real-name + content-moderation controls
  6. 6.Personal-information protection (PIPL) alignment
  7. 7.Incident + illegal-content handling
Next 7 dates & movements
  • ongoingAI-generated-content labelling rules in effect + enforced
  • 2026New TC260 national standards on GenAI security published
  • ongoingAlgorithm-filing enforcement + provider registration
  • ongoingData-export + PIPL interplay for AI services
  • 2026–27Sectoral rules (finance, health, autonomous) expand
  • ongoingDeep-synthesis + deepfake provisions enforced
  • ongoingCross-border AI service restrictions evolve
How CSOAI covers it

Filing/registration checklist, security self-assessment, and content-labelling controls mapped to the same evidence spine as EU/US.

Open in the Hive →Govern a case →
UK AI Regulation
United Kingdom · DSIT · sector regulators (ICO, FCA, CMA, Ofcom)
Status shifting

A principles-based, pro-innovation approach delivered through existing regulators, with a possible AI bill for the most capable models under active debate.

Top 7 tools needed
  1. 1.Cross-sector principle mapping (safety, transparency, fairness, accountability, contestability)
  2. 2.Regulator-specific compliance (ICO/FCA/CMA/Ofcom)
  3. 3.AI assurance + audit techniques
  4. 4.Algorithmic transparency records (ATRS for public sector)
  5. 5.DPIA / data-protection alignment (UK GDPR)
  6. 6.Model risk + governance documentation
  7. 7.Incident + harm reporting
Next 7 dates & movements
  • 2026Direction on a UK AI bill for frontier models — status shifting, verify
  • ongoingAI Safety Institute evaluations of frontier models
  • ongoingSector regulators publishing AI strategic approaches
  • 2026Algorithmic Transparency Recording Standard scaling in public sector
  • ongoingICO guidance on AI + data protection updated
  • ongoingInternational interoperability (EU/US) positioning
  • TBDStatutory footing decisions for the AI Safety Institute
How CSOAI covers it

The five cross-sector principles as a live checklist, regulator mapping, and assurance evidence — bridged to EU/ISO so UK-first orgs stay portable.

Open in the Hive →Govern a case →
Canada AIDA / AI policy
Canada · ISED · Office of the AI & Data Commissioner (proposed)
Status shifting

The Artificial Intelligence and Data Act (part of Bill C-27) did not pass before Parliament was prorogued; Canada's statutory AI direction is being reset — track the voluntary code meanwhile.

Top 7 tools needed
  1. 1.Voluntary Code of Conduct (generative AI) alignment
  2. 2.High-impact-system identification
  3. 3.Risk mitigation + monitoring plan
  4. 4.Transparency + disclosure records
  5. 5.Human oversight measures
  6. 6.Bias + harm assessment
  7. 7.PIPEDA / privacy alignment
Next 7 dates & movements
  • 2026New legislative direction post-C-27 — status shifting, verify
  • ongoingVoluntary Code of Conduct adoption by developers
  • ongoingProvincial (Quebec Law 25) privacy interplay
  • 2026Federal consultation on a revised AI framework
  • ongoingAlignment signalling with EU AI Act + NIST
  • TBDStanding up an AI & Data Commissioner function
  • ongoingPublic-sector AI directive updates
How CSOAI covers it

Voluntary-Code alignment now, structured so you flip to statutory obligations the moment Canada's framework lands — no rework.

Open in the Hive →Govern a case →
Singapore Model AI Governance
Singapore · IMDA · PDPC
Voluntary

A practical, voluntary framework — the Model AI Governance Framework (incl. Generative AI) plus AI Verify testing toolkit — widely used as an implementation template across APAC.

Top 7 tools needed
  1. 1.Model AI Governance Framework mapping
  2. 2.AI Verify testing + reporting toolkit
  3. 3.Risk-based deployment controls
  4. 4.Data governance + PDPA alignment
  5. 5.Human-in-the-loop design patterns
  6. 6.Transparency + stakeholder communication
  7. 7.Incident management
Next 7 dates & movements
  • ongoingAI Verify + GenAI evaluation sandbox expansion
  • 2026Updated GenAI governance guidance + testing standards
  • ongoingCross-border interoperability work (ASEAN + global)
  • ongoingSectoral guidelines (finance/health) from MAS + others
  • 2026–27AI Verify Foundation tooling ecosystem grows
  • ongoingAlignment with ISO 42001 for certification-minded firms
  • ongoingPublic-sector AI adoption playbooks
How CSOAI covers it

Model Framework mapping + AI Verify-style testing evidence, bridged to ISO 42001 so a Singapore deployment ports globally.

Open in the Hive →Govern a case →
Dates and obligations are compiled from established framework knowledge and evolve constantly — always verify against the primary regulator before you rely on a specific date. The Council assistant's live read pulls the current picture; volatile items are marked Status shifting.
Comply once. Crosswalk everywhere.

One evidence set, mapped across every regime above — signed to Layer 0, provable not promised. That's the difference between a governance program and a pile of PDFs.