← Back to AI Governance

● One-day training program

AI Governance for Directors & Senior Management

A one-day, operationally-focused program for management-level "Directors" who need to build and run an AI governance program — not pass a certification exam. Condensed from the IAPP AIGP Body of Knowledge v2.1 (58 lessons, 4 domains) into what leaders actually need to act on: accountability, the regulatory landscape, lifecycle controls, and vendor/deployment oversight.

6.5 hrs
of programming (9:00–17:00)
4
modules, mapped to the 4 governance domains
1
take-home oversight toolkit
Cross-industry
general applicability

How this day is designed

The source material is written for an individual studying to pass a certification exam — dense, recall-oriented, organized around what gets tested. A room of senior managers needs the opposite: a small number of decisions and questions they can act on Monday morning. This program keeps the four-domain structure (it's a sound map of the discipline) but re-weights every session from "know this for a test" to "here's what to build, and what to ask your team."

Full-Day Agenda

A standard 9:00–17:00 day: 6.5 hours of programming, a working lunch, and two short breaks. Every block below can compress or stretch by ±10 minutes without breaking the flow — see Facilitator Notes for how to adapt it to a half-day or a two-day format.

09:00 – 09:20
Welcome & Why This Is a Management Issue Now
KickoffFraming: what makes AI different to govern (complexity, opacity, autonomy, scale), and what "good" looks like by the end of the day.
09:20 – 10:30
Module 1 — Foundations, Principles & Accountability
Presentation + DiscussionWhy AI needs governance, the responsible-AI principles, and — the core of the module — who is actually accountable when something goes wrong.
10:30 – 10:45
Break
10:45 – 12:15
Module 2 — The Regulatory Landscape You Operate Under
Presentation + ExerciseEU AI Act risk tiers and obligations, NIST AI RMF, and a fast tour of the US and global patchwork — enough to know what applies and when to call counsel.
12:15 – 13:15
Lunch
13:15 – 14:30
Module 3 — Governing the Build: Lifecycle, Assessments & Bias
Presentation + DiscussionThe lifecycle policy gates, impact assessments (DPIA/FRIA/AIA), the four TEVV testing questions, and the eight-bias taxonomy for reading incident reports.
14:30 – 14:45
Break
14:45 – 16:00
Module 4 — Governing Deployment: Vendors, Monitoring & Incidents
Presentation + ExerciseDeploy-decision factors, the vendor/licensing contract checklist, ongoing monitoring duties (yes, even for tools you buy), and incident documentation.
16:00 – 16:45
Applied Session — Management Oversight Toolkit & Case Discussion
WorkshopAttendees apply the day's tables to one real system from their own organization; walk out with a filled-in oversight checklist.
16:45 – 17:00
Wrap-Up, Action Plan & Q&A

By the end of the day, attendees will be able to

Domain I, re-leveledModule 1 — Foundations, Principles & Accountability

The shortest module in scope but the one everything else depends on. Get accountability right here and the rest of the day is implementation detail; get it wrong and no framework in Module 2 will save you.

Why AI needs its own governance

~15 min

Objective: agree on the vocabulary and the risk rationale before assigning any accountability.

An AI system is a machine-based system that infers from the inputs it receives how to generate outputs — predictions, content, recommendations or decisions. What makes it a distinct governance problem, versus ordinary software, is a specific combination of traits: complexity (hard to fully audit), opacity (decision logic isn't always inspectable), autonomy (can act with less human review), and speed and scale (one flawed model can affect millions of decisions before anyone notices). Every control in this document exists to answer one of those four traits.

The responsible-AI principles — and what they look like in practice

~15 min

The same eight principles recur across every framework covered today (OECD, NIST, the EU AI Act, most corporate charters). The table is the point: a principle only governs once it becomes a requirement with an owner.

PrincipleWhat it meansWhat it looks like in practice
EthicsRespect human rights and societal values — the umbrella over the restEthics review; clear red lines for unacceptable uses
FairnessAvoid unjust bias; comparable treatmentA chosen fairness metric, a threshold, a disparate-impact test
Safety & reliabilityPerforms as intended within safe limitsRobustness testing, guardrails, fail-safes
PrivacyLawful, minimal use of personal dataData minimization, DPIAs, retention limits
SecurityProtected from attack and manipulationThreat modeling, access control, poisoning defenses
Transparency & explainabilityPeople know it's AI and can understand outputsDisclosure, model cards, audience-appropriate explanations
AccountabilityNamed people answer for outcomesA single accountable owner; an audit trail
Human-centricityRespects human autonomy and wellbeingHuman-in-the-loop, opt-outs, meaningful redress

These principles routinely pull against each other — more transparency can expose personal data or invite gaming; tighter fairness constraints can cost raw accuracy; heavy human oversight can slow throughput. Governing well means naming the trade-off explicitly and recording who signed off on the balance, not pretending it doesn't exist.

Discussion A hiring platform commits to "fairness." Applying that concretely means choosing a fairness metric, testing for disparate impact, and accepting that publishing the model's logic (more transparency) may trade against gaming and security. Who in your organization is authorized to sign off on that trade-off today — and if the honest answer is "no one," what does that tell you?

Who is accountable — RACI, the Three Lines, and the lifecycle

~30 min

Objective: leave with a concrete accountability structure, not just a principle that "everyone owns AI."

Effective AI governance is distributed but accountable. The goal is to avoid both failure modes: a single overwhelmed owner, and diffuse responsibility where "everyone" owns AI and therefore no one does. Two tools do the work:

  • RACI (Responsible, Accountable, Consulted, Informed) mapped per lifecycle stage — one name in the Accountable column for every system, even when many functions are Consulted.
  • The Three Lines model. First line — the business/operational functions that own and manage the risk day-to-day (the teams that build and run the AI). Second line — risk management and compliance, which sets policy and oversees the first line, but doesn't own the risk itself. Third line — internal audit, giving independent, objective assurance to senior management and the board that the first two lines are actually working. (There is no fourth line — external auditors and regulators sit outside the model.)
Who leads at each lifecycle stage — use this as the skeleton of your RACI
StageWho leadsWhat they own
Planning & designExec sponsor, product owner, governance/ethics leadSet acceptable-use limits, screen ethical risk, fund the build
Data collection & prepData engineering, legal, privacy/complianceConfirm lawful data rights, check representativeness & bias, apply privacy controls
Model development & trainingData scientists, ML engineers, risk partnersTrain and validate, record design assumptions, measure bias & performance
Validation & testingQA, domain experts, complianceRun robustness & security tests, confirm accuracy & reliability
Verification & releaseGovernance board, accountable executiveReview pre-deployment evidence, decide go/no-go, accept or remediate residual risk
Deployment & integrationAI/ML ops, IT, business ownerStand up monitoring & alerts, hold to service levels, control the rollout
Operation & monitoringRisk owners, end users, complianceWatch for drift, gather feedback, run incident response, keep documentation current
Decommissioning / retirementLegal, IT, governance committeeRetire models safely, preserve logs & records, update the inventory
Red flags to check for right now
  • Diffuse ownership. If "everyone" is responsible for AI, accountability evaporates the moment something goes wrong.
  • IT-only governance. Excluding legal, ethics and the affected business unit creates blind spots in exactly the highest-risk areas.
  • Principles as wall art. Publishing values without operationalizing them into requirements and tests is governance theater.
Management takeaway Five actions to charter this month: (1) publish a RACI per lifecycle stage, (2) assign one accountable owner per system — not a committee, (3) charter a cross-functional AI governance committee with real decision rights, (4) adopt a three-lines structure so oversight and assurance are separated from ownership, (5) secure a named board or C-suite sponsor so the program has budget and authority.

Policies that follow the system end to end

~10 min

A policy only governs if it is an enforceable gate — explicit entry/exit criteria a system must clear to advance, with evidence to prove it. "No approved use-case assessment, no data work; no test sign-off, no release."

Eight lifecycle stages, each with its policy gate
Lifecycle stagePolicy gateWhat it requires to pass
Use-case assessmentIntake approvalDefined purpose and value, plus an initial risk screen
Risk & impactImpact-assessment sign-offDocumented impacts and mitigations
Data acquisition & useData-use approvalLawful rights, quality and representativeness checks
Development & trainingBuild reviewDesign assumptions and risk controls recorded
Testing & validationTest sign-offValidation, robustness, security and bias results met
DeploymentRelease gateConformity, oversight, rollback and a monitoring plan
MonitoringPeriodic re-reviewDrift, performance and incidents reviewed on a cadence
Incident managementResponse readinessDetection, escalation and disclosure paths defined

It's a loop, not a line: monitoring, incidents and drift feed back into earlier stages, triggering re-assessment or even redesign. Live systems get re-reviewed on a cadence, not approved once and forgotten.

Domain II, re-leveledModule 2 — The Regulatory Landscape You Operate Under

The goal here is not legal fluency — it's pattern recognition. By the end of this module, attendees should be able to tell which of the four "kinds" of framework they're looking at, and flag when a system likely needs legal review.

Four frameworks, four different jobs

~15 min

The single most useful mental model for the regulatory landscape: these frameworks are not competing versions of the same thing. They sit at different altitudes and do different jobs.

FrameworkWhat it isBinding?
OECD AI PrinciplesThe values — the first intergovernmental AI standard (2019, updated 2024)Soft law — non-binding, no penalties
NIST AI RMFThe process — a voluntary US risk-management operating modelVoluntary — no penalties, but widely adopted
ISO/IEC 42001The certificate — an AI management system standard you can be audited againstVoluntary but certifiable
EU AI ActThe law — a binding, risk-tiered regulation with finesBinding — fines up to €35M / 7% of global turnover

The OECD's technology-neutral definition of an AI system — a system that infers from inputs how to generate outputs — is the definition the EU AI Act itself builds on. Know that lineage and the rest of this section falls into place.

The EU AI Act, consolidated

~40 min

Objective: correctly sort a described system into one of four risk tiers, and know what obligation follows.

The world's first comprehensive, horizontal AI law. It is risk-based (obligations scale with risk tier), extraterritorial (applies to non-EU providers/deployers when the system's output is used in the EU, much like the GDPR), role-based, and technology-neutral.

The risk pyramid

Prohibited
Unacceptable risk — banned outright
High-risk
Strict obligations & conformity assessment
Limited risk
Transparency / disclosure duties
Minimal risk
No specific obligations — e.g. spam filters, game AI

Prohibited practices — banned outright

  • Harmful manipulation and subliminal or deceptive techniques
  • Exploiting vulnerabilities of age, disability or social/economic situation
  • Social scoring by or for public authorities leading to detrimental treatment
  • Untargeted scraping of facial images to build recognition databases
  • Emotion recognition in the workplace and in education (narrow exceptions)
  • Biometric categorization inferring race, beliefs, or sexual orientation
  • Certain predictive policing based solely on profiling
  • Real-time remote biometric identification in public spaces by law enforcement (narrow exceptions)

High-risk: the systems that surprise people

A system is high-risk if it's a safety component of a regulated product, or falls in an Annex III use area: biometrics, critical infrastructure, education & vocational training (admissions, scoring, proctoring), employment (recruitment, screening, promotion, termination), essential services including credit scoring and insurance pricing, law enforcement/migration, and administration of justice. Education, employment and credit are the classic blind spots — they sound routine but carry full high-risk obligations.

Who does what — obligations follow the role
RoleCore duties
ProviderThe heaviest duties: risk management, data governance, documentation, conformity assessment, CE marking, registration, post-market monitoring
DeployerUse per instructions, ensure human oversight, monitor, keep logs, inform affected people, run a FRIA for certain high-risk uses
ImporterVerify the provider's conformity assessment, documentation and CE marking before placing on the EU market
DistributorCheck that marking and documentation are present before making a system available

Watch this one: a deployer who substantially modifies or rebrands a system becomes a provider — and inherits the heavier duties.

General-purpose AI (GPAI) — a separate regime

All GPAI providers must maintain technical documentation, publish a summary of training data used, and put in place a copyright policy. GPAI with systemic risk — presumed above a 10²⁵ FLOP training-compute threshold — carries extra duties: model evaluation, adversarial testing, systemic-risk assessment, serious-incident reporting and cybersecurity.

Penalties and timeline

BreachMaximum fine
Prohibited-practice breaches€35M or 7% of global annual turnover, whichever is higher
Most other obligation breaches (high-risk, GPAI, transparency)€15M or 3%
Supplying incorrect/misleading information to authorities€7.5M or 1.5%
DateWhat applies
Aug 2024Entry into force of the Regulation
Feb 2025Prohibited practices and AI-literacy obligations apply
Aug 2025GPAI obligations, governance bodies and penalties apply
Aug 2026Most high-risk (Annex III) obligations apply
Aug 2027High-risk systems that are safety components of regulated products apply
Exercise — which tier is this? Run three of your own systems (or these examples: a resume-screening tool; an internal chatbot that answers HR policy questions; a spam filter) through the pyramid as a group. For each: which tier, which role are you (provider or deployer), and what's the one obligation you're least confident your organization already meets?

NIST AI Risk Management Framework

~15 min

The framework to reach for wherever no AI-specific law yet applies — voluntary, sector-agnostic, and paired with a companion Playbook of suggested actions.

FunctionWhat it does
Govern (cross-cutting)Cultivate a culture of risk management — policies, accountability, roles and oversight that make the other three functions work
MapEstablish context and frame the risks of the system and its use case
MeasureAnalyze, assess and benchmark the identified risks (bias, robustness, security)
ManagePrioritize, respond to and monitor the risks, allocating resources to the greatest

Pair each function with the seven trustworthy-AI characteristics it's meant to produce: valid & reliable, safe, secure & resilient, accountable & transparent, explainable & interpretable, privacy-enhanced, and fair (harmful bias managed).

The wider patchwork — enough to know when to call counsel

~15 min

There is no single US federal AI law. Instead: a growing state patchwork, plus existing federal regulators applying old authority to new technology.

Law / regulatorWhat it does
Colorado AI ActFirst comprehensive US state AI law — duty of care against algorithmic discrimination in consequential decisions
NYC Local Law 144Requires an independent bias audit before using an automated hiring/promotion tool, plus candidate notice
Illinois BIPAStrict consent rules for biometric identifiers (face, fingerprint, voice) — with a private right to sue
California AB 2013 / SB 942 / SB 1001Training-data transparency / AI-content provenance & labeling / "a bot must say it's a bot"
FTC / EEOCApply existing authority to AI — overstated "AI-washing" claims and AI hiring tools that produce discriminatory outcomes

Globally: the EU AI Act (binding law), South Korea's AI Basic Act (2nd comprehensive national law, effective Jan 2026), the Council of Europe Framework Convention (the first binding AI treaty), China's vertical rules and algorithm registry, and softer-touch principles in the UK, Singapore and Japan. Takeaway for management: if a system touches hiring, credit, biometrics, or EU users, get legal involved early — this is not a landscape to navigate on instinct.

Domain III, re-leveledModule 3 — Governing the Build: Lifecycle, Assessments & Bias

This is the build half of the lifecycle. The goal is to leave with the vocabulary to ask sharp questions of a technical team — not to become the technical team.

The AI lifecycle, end to end

~15 min

Governance is not a one-time gate — it follows the system from first idea to retirement, and it's a loop, not a line: monitoring, incidents and drift feed back into earlier stages. Four things run continuously across every stage: governance & accountability, risk management, documentation, and human oversight.

Build half (this module): use case → impact assessment → data governance → train → test (TEVV) → release readiness. Deploy half (Module 4): select model → deploy → monitor → maintain/retrain → deactivate.

Impact assessments — know which one applies

~15 min

An AI impact assessment systematically evaluates potential harms to individuals, groups and society, the proportionality of the system, and the mitigations — ideally starting before development and iterating as the design firms up. Several overlapping assessments exist:

AssessmentCovers
DPIAPrivacy / data-protection risk
FRIAFundamental-rights impact (required of certain EU AI Act deployers)
AIAThe whole AI risk picture — harms, proportionality, mitigations
ISO/IEC 42005A recognized method for structuring an AI impact assessment
Red flags
  • Assessment as a launch formality. Done once at the end, it cannot shape design — assessments must start early and iterate.
  • Ignoring less-intrusive alternatives. Proportionality requires asking whether a lighter approach achieves the same goal.

TEVV — the four questions a test report is actually answering

~15 min

Four testing activities get blurred together in practice. Knowing which question each answers is enough to read any technical test report critically.

ComponentThe question it answersMeasured against
TestDoes this specific function work under known conditions?Predefined inputs & expected outputs
EvaluateHow good is the whole system, overall?Metrics across the battery — accuracy, robustness, fairness
VerifyDid we build it right?Design specifications & requirements
ValidateDid we build the right thing?Real-world need & intended use

The most useful distinction for a manager: Verify asks "did we meet the spec," Validate asks "does this actually solve the real problem." A system can pass verification and still fail validation.

The eight-bias taxonomy — for reading an incident, not passing an exam

~20 min

Each type of bias enters the system at a different point. Knowing which point helps you ask the right follow-up question when a team reports a fairness issue instead of accepting "it's biased" as a full explanation.

BiasEnters at…What's actually wrong
HistoricalReal-world inequity, accurately recordedThe data is correct, but the world it records was unfair
RepresentationSampling — who is in the dataSome groups are under-represented vs. the deployment population
MeasurementFeatures & labels (proxies)The proxy is flawed, or measured differently across groups
AggregationModel designOne model is forced onto distinct subgroups that behave differently
EvaluationThe test set / benchmarkThe benchmark is unrepresentative, so headline metrics mislead
DeploymentContext of useUsed outside the context it was built or validated for
AutomationHuman behavior (over-trust)People over-rely on outputs and stop applying judgment
ConfirmationHuman cognition (belief)People favor outputs that confirm what they already expected

The classic mix-up: historical vs. representation vs. measurement all look like "biased data" in a post-mortem — the difference is why. And two of the eight (automation, confirmation) aren't data problems at all — they're human behavior, and the fix is oversight design and training, not re-sampling a dataset.

Discussion scenario A résumé screener trained on a decade of the company's own hiring decisions historically favored one group for a given role. The data is accurately recorded — no errors. Which bias is this, and who in the room would have caught it before launch?
(Answer: historical bias — accurate data, unfair world. Not representation bias, since no group is under-sampled; not measurement bias, since the labels aren't a flawed proxy.)

Domain IV, re-leveledModule 4 — Governing Deployment: Vendors, Monitoring & Incidents

Most organizations are deployers of someone else's model far more often than they are builders of their own. This module is deliberately weighted toward that reality.

Before you deploy: context, model type, deployment options

~10 min

Three questions to settle before committing to a system: the context of the use case (who's affected, how consequential), the model type (classic predictive vs. generative vs. agentic carry different risk profiles), and the deployment option (build in-house, license, or embed a third-party API — each shifts where obligations and liability sit).

The vendor contract is a primary control

~25 min

Objective: leave able to hand this table to procurement or legal before the next AI vendor contract is signed.

When you're buying rather than building, the contract is your governance control. Default vendor terms are written to protect the vendor.

Eight clauses to scrutinize in any AI vendor or licensing agreement
TermWhat to check
Data useWill your inputs train the vendor's model?
IP & ownershipWho owns the outputs?
Performance / SLACommitted accuracy, uptime and support
Liability & indemnityWho bears IP-infringement and harm risk?
Transparency & auditRights to documentation, bias evidence, audit
Security & sub-processorsControls, and the vendor's own downstream suppliers
Model-change noticeNotification when the model materially changes
Exit & portabilityGet your data/config out; avoid lock-in
Red flags
  • Click-through acceptance of AI terms. Default terms may grant training rights over your data and disclaim all liability.
  • No audit or incident-notification rights. This leaves you blind and unprotected when the vendor's AI fails.

Monitoring doesn't stop because you bought it instead of building it

~20 min

The vendor's benchmarks do not capture your data and your population — only ongoing monitoring in your own production environment does.

Watch forWhy
DriftYour data/population diverges from the lab's
PerformanceAccuracy can decay in your specific context
BiasFairness must hold for your population, not the vendor's test set
OutagesOperational health and availability
Vendor updatesValidate changes before you adopt them, don't inherit them blind
Red flags
  • Assuming the vendor monitors for you. You must monitor performance and bias in your own context regardless of what the vendor reports.
  • Adopting vendor updates blindly. Unvalidated model changes can break or bias your use case without warning.

Incidents, documentation, and eventually turning it off

~20 min

Deployers keep records too — an incident log, an issue/risk register, and a post-market monitoring plan, feeding the provider and regulators where required. Reporting runs both ways: you often must notify the provider, who has their own downstream duties.

RecordPurpose
Incident logEvidence, learning and reporting
Issue / risk registerTrack open items with named owners
Post-market monitoring planTrack real-world performance over time
Provider / regulator reportsMeet two-way notification duties

And finally: every system needs a designed way to be turned off. Be able to deactivate, localize, or retire a system safely — with a fallback process and a plan for the data it leaves behind.

Red flags
  • No deployment incident records. Without them you cannot report, learn, or demonstrate diligence after the fact.
  • Not notifying the provider. This breaks the post-market loop and may breach legal duties.

Applied Session — Management Oversight Toolkit

16:00–16:45. Attendees pick one real AI system from their own organization and work through this checklist live, in small groups. The goal is a filled-in page, not a perfect one — gaps found today are cheaper than gaps found in an incident review.

Questions to ask your team about any AI system

Quick-reference: tell the frameworks apart

If someone says…They probably mean…
"the values-based principles"OECD AI Principles (soft law, 2019/2024)
"the risk-management process"NIST AI RMF — Govern · Map · Measure · Manage (voluntary, US)
"the certification / audit standard"ISO/IEC 42001 (certifiable AI management system)
"the law with the fines"EU AI Act (binding, up to €35M / 7% turnover)

30 / 60 / 90-day follow-up plan

By day…Do this
30Inventory every AI system in use (including "shadow AI" tools teams adopted on their own) and assign one accountable owner to each
60Run the risk-tier and vendor-contract checklists above against your three highest-exposure systems; escalate gaps to legal/compliance
90Charter the AI governance committee, publish the lifecycle RACI, and schedule the first periodic re-review cadence

Quick-Reference Glossary

The ~25 terms used most often across today's session. For the complete 281-term glossary behind this program, see the full course extract referenced in Facilitator Notes.

AI system
A machine-based system that infers from inputs how to generate outputs — predictions, content, recommendations or decisions.
Provider
Builds or markets an AI system — carries the heaviest EU AI Act duties.
Deployer
Uses an AI system under its own authority; becomes a provider if it substantially modifies or rebrands the system.
GPAI
General-purpose AI model — a separate EU AI Act regime, with extra duties above a systemic-risk compute threshold.
High-risk system
An EU AI Act tier carrying strict obligations and conformity assessment — includes employment, education and credit use cases.
RACI
A matrix assigning who is Responsible, Accountable, Consulted and Informed for each activity.
Three lines model
Business owns risk (1st line); risk/compliance sets policy and oversees (2nd line); internal audit gives independent assurance to the board (3rd line).
DPIA
Data protection impact assessment — covers privacy risk from high-risk personal-data processing.
FRIA
Fundamental-rights impact assessment — required of certain high-risk EU AI Act deployers.
AIA
AI impact assessment — the whole-picture evaluation of harms, proportionality and mitigations.
TEVV
Test, Evaluation, Verification, Validation — the four-part testing battery, each answering a different question.
Model card
A standard document describing a model's purpose, data, performance and limits.
Drift
Degradation as live inputs or relationships diverge from what the model was trained on.
Disparate impact
A neutral-looking practice that nonetheless disadvantages a protected group — a key fairness test.
SLA
Service-level agreement — committed performance and availability terms in a vendor contract.
Indemnification
A contractual shift of liability — for example, for IP infringement caused by a vendor's model.
Human-in-the-loop
A design requiring a person to review, confirm or be able to override an AI decision.
Post-market monitoring plan
A plan to track a deployed system's real-world performance after launch.
Stage gate
A checkpoint a system must clear — with evidence — before moving to the next lifecycle stage.
Go/no-go decision
A deliberate release decision to proceed or halt, based on whether gate criteria are met.
Conformity assessment
The formal EU AI Act process a high-risk provider completes before a system reaches market.
Soft law
Influential but non-binding rules (e.g. OECD Principles) — no penalties attach to non-compliance.
Proportionality
Whether a system's benefits justify its risk and intrusiveness, including whether a lighter approach would do.
Accountable owner
The single named person answerable for a given AI system or decision.
AI governance committee
A cross-functional body that sets policy, reviews high-risk use cases, and oversees the program.

Facilitator Notes

Adapting the timing

Materials to prepare

Provenance of this material

This program condenses and re-levels content from a self-study course built on the IAPP AIGP Body of Knowledge v2.1 (effective Feb 2026) — a professional-certification syllabus, not an official IAPP training product. It has been substantially rewritten here: exam-recall framing removed, self-test Q&A converted to management discussion prompts, and content re-weighted toward operational, cross-industry, management-level application rather than certification recall. Verify current thresholds, fine amounts and effective dates against the primary sources below before publishing this externally or relying on it for compliance decisions — regulatory details (especially EU AI Act phase-in dates and US state law) move quickly.

Primary sources for further reading

kashifz.com