Regulations & Standards

Know what your board is accountable for.

AI oversight has become a board-level duty. This is the landscape of obligations, standards, and exposure, in board language, not legalese.

Updated November 2026

Start from the question in front of your board. The map shows which regimes apply and what each asks of you.

In forceComing into forceVoluntary or guidanceProposed
European UnionAI Act: transparency in force; high-risk duties from Dec 2027.
United States · FederalNo AI statute. SEC, FTC and executive orders; preemption not enacted.
United States · StatesColorado and Illinois from Jan 2027; California and Texas in force.
United KingdomPrinciples enforced by sector regulators.
CanadaNo federal AI law; privacy law applies.
South KoreaAI Basic Act.
ChinaGenerative AI, filing and labelling rules.
JapanAI Promotion Act; light touch.
SingaporeVoluntary frameworks; finance guidance.
BrazilRisk-based AI bill in progress.
StandardsNIST AI RMF, ISO/IEC 42001, OECD principles.

Summaries, not legal advice. Check the current text with counsel before relying on any of them.

The landscape

Vela Board AdvisorsAI Rules Landscape

35 rules and standards, one picture.

Each circle is a law, rule or standard, placed by when it took or takes effect. Bigger circles matter more to boards. Follow the story, or click a region.

Vela tally of instruments tracked on this page · Size = board relevance · Dates approximate · Not legal advicevelaboards.com/regulations

Timeline

Vela Board AdvisorsAI Regulation Timeline

The dates your board should plan around.

EUUS statesCaliforniaAlready in force
2026202720282029
Today
Jan 2026California & TexasADMT rules and TRAIGACalifornia
Aug 2026EU transparencyChatbot and deepfake dutiesEU
Dec 2026EU content markingAI content must be markedEU
Jan 2027Colorado & IllinoisAutomated decisions; 72-hour clockUS states
Dec 2027EU high-riskHiring, credit, educationEU
Jan 2028AI evaluatorsCalifornia frameworkCalifornia
Aug 2028EU regulated productsAI inside regulated productsEU
Jan 2029Auditor registryCalifornia AI auditorsCalifornia
  1. Jan 2026California & TexasADMT rules and TRAIGA in force
  2. Aug 2026EU transparencyChatbot and deepfake duties
  3. Dec 2026EU content markingAI content must be marked
  4. Jan 2027Colorado & IllinoisAutomated decisions; 72-hour clock
  5. Dec 2027EU high-riskHiring, credit, education
  6. Jan 2028AI evaluatorsCalifornia framework
  7. Aug 2028EU regulated productsAI inside regulated products
  8. Jan 2029Auditor registryCalifornia AI auditors
Key AI regulation dates for boards · Updated November 2026 · Not legal advicevelaboards.com/regulations
Share on LinkedIn

How it fits together

Four layers of accountability.

YOUR DUTIESBINDING LAWSTANDARDSON THE HORIZONYourboard
  1. 1
    Your dutiesCare · loyalty · oversightAlready apply to AI. The record shows whether you met them.
  2. 2
    Binding lawEU AI Act · US states · SEC · FTC · Korea · ChinaArrives by jurisdiction, sector and use case.
  3. 3
    StandardsNIST AI RMF · ISO/IEC 42001 · OECDVoluntary, but the benchmark regulators, insurers and courts use.
  4. 4
    On the horizonPreemption · Brazil · CanadaProposed. Track it; don’t build for it yet.

What the rules are about

Vela Board AdvisorsAI Rules by Topic

Transparency and governance lead. Incidents are rising.

Each block is sized by how many of the instruments on this page address that topic. Click a block to see which ones.

Click a block
Vela tally · An instrument can cover several topics · Not legal advicevelaboards.com/regulations

Tracker

Who it binds, and how much it matters.

InstrumentWho it bindsStatusPenaltiesBoard relevance
EU AI ActProviders and deployers in or selling into the EUComingYes
SEC & FTCPublic companies; anyone making AI claimsIn forceYes
Colorado SB 26-189Developers and deployers of automated decisionsComingYes
Illinois SB 315Large frontier model developersComingYes
California ADMT rulesBusinesses using AI for significant decisionsIn forceYes
Texas TRAIGADevelopers and deployers in TexasIn forceYes
NYC Local Law 144Employers using AI in hiringIn forceYes
NIST AI RMFAnyone; cited in contracts and policyVoluntaryNo
ISO/IEC 42001Anyone seeking certificationVoluntaryNo

Who is watching

Vela Board AdvisorsScrutiny Map

Who will look at your AI oversight.

Regulators are only part of it. Hover to see what each one looks at, or filter by type.

Illustrative, not exhaustive · Not legal advicevelaboards.com/regulations

Your exposure

How many rules reach your company?

Choose your profile. The bars count rules and standards from our tracker that would reach a company like yours, by region.

Company
Where you operate
Your AI touches
0rules and standards reach you

    An illustrative selection from our tracker, not an exhaustive legal map. Applicability depends on facts; confirm with counsel.

    Your role

    Which duties are yours depends on the role you play.

    Most companies play more than one. Solid blocks are in force today; striped blocks are coming.

    EU AI Act high-risk conformityColorado AI Act developer dutiesEU AI Act general-purpose model dutiesCalifornia frontier AI transparency (SB 53)
    DeveloperYou build or train the model
    EU AI Act high-risk deployer dutiesColorado AI Act deployer dutiesEU AI Act prohibited practices and AI literacyCalifornia automated decision-making rulesTexas Responsible AI Governance Act
    DeployerYou use AI in your operations
    SEC examination focus on AISEC risk-factor and incident disclosureSEC and FTC action on overstated AI claimsDelaware oversight duty (Caremark)
    Public companyYour shares are listed
    EU AI Act: employment is high-riskNew York City bias audits (Local Law 144)Illinois AI in employment lawExisting anti-discrimination law
    EmployerAI touches hiring or staff
    In forceComing

    For you, personally

    Your protection depends on where you sit.

    United StatesUnited KingdomEU (e.g. Germany)CanadaAustraliaCourts defer to business judgmentYesPartlyYesYesYesDuty of care set in statuteNoYesYesYesYesCharter can exclude liability for careYesNoNoNoNoCompany can indemnify directorsYesPartlyPartlyYesPartlyD&O insurance availableYesYesPartlyYesYesPersonal regulatory accountabilityPartlyYesYesPartlyYes
    Gets through almost everywhereBad faithBreach of loyaltyPersonal profitFraudIgnored red flags

    A simplified comparison. Protection varies by state or country, company documents and policy wording. Vela does not provide legal advice; confirm with counsel.

    Compare regions

    Same technology. Different expectations.

    Pick two regions to compare how far each goes on six dimensions.

    compared with

    Vela's simplified reading of each region's binding rules, scored 1 to 5. Illustrative only.

    Standards

    The benchmarks you'll be measured against.

    Voluntary guidanceCertifiable and auditablePrinciplesOperational detailSet the valuesProve it to othersRun it day to day
    NIST AI RMFFramework · US reference
    ISO/IEC 42001Certifiable management system
    ISO/IEC 23894Risk management guidance
    OECD AI PrinciplesValues adopted by 47+ countries
    NIST · NSA · CISA agent practicesTen practices for agent incidents

    Exposure

    Where boards are exposed.

    • Securities & disclosureOverstated AI claims; undisclosed incidents
    • DiscriminationHiring, lending and pricing models
    • IP & copyrightTraining data and output ownership
    Six fault linesThe same capability that creates value concentrates liability.
    • Privacy & dataBiometrics and cross-border flows
    • Vendors & agentsThird-party models and autonomous agents
    • Safety & productAI-enabled products that cause harm

    Inherited exposure

    Your vendors’ AI becomes your exposure.

    Most of your AI is built from other companies’ parts. Regulators and courts still hold your company accountable for all of it. Each layer raises a question the board should be able to answer.

    Your companyAccountable for everything below, wherever it came from
    Outsourced processesWhere suppliers use AI on your customers
    AskWho checks the AI they use on our customers?
    Agents acting for youSoftware that takes actions, not just answers
    AskWhat can they do without a human signing off?
    Software with built-in AITools you already use, updated with AI
    AskDid anyone approve the AI that arrived in the last update?
    Model providersThe models behind your AI features
    AskCan they change the model without telling us?
    Data suppliersThe data models are trained and run on
    AskDo we have the rights to use the data?
    Cloud platformsWhere it all runs
    AskWhat happens if access is cut off?

    The incident clock

    From AI incident to boardroom consequence.

    Each step narrows the time the board has to act, and each leaves a record.

    The incidentDay 0

    A model fails, leaks data or misleads customers.

    The clock startsHours

    Materiality and notification assessments begin.

    DisclosureDays

    Public companies: Form 8-K within 4 business days of a material cybersecurity determination. GDPR breaches: 72 hours to the regulator.

    RegulatorsWeeks

    EU AI Act serious incidents: up to 15 days for providers of high-risk systems, shorter for the most severe.

    LitigationMonths

    Shareholder and consumer claims often follow a stock drop or disclosure.

    D&O and insuranceThroughout

    Notice obligations and exclusions decide what is covered.

    Timelines summarised from SEC Form 8-K Item 1.05, GDPR Article 33 and EU AI Act Article 73. Requirements depend on the facts; confirm with counsel.

    In court

    What helps and what hurts when oversight is tested.

    Courts rarely second-guess a decision. They look at whether the board paid attention.

    Lowers exposureRaises exposure
    Documented questions and challenge in minutes
    Ignored red flags
    Named committee owner with a reporting cadence
    Overstated AI claims to investors
    Independent testing or outside advice
    Incidents not escalated or disclosed
    Rehearsed escalation path
    AI never on the board agenda
    Director education on record
    No owner for AI oversight

    Relative weight is Vela's illustrative reading of Delaware oversight cases and enforcement trends. Not legal advice.

    On the record

    What "good" looks like on the record.

    Board Minutes · ExtractItem 6 · AI underwriting

    AI oversight remains with the Risk Committee, chaired by Ms Ellis, who owns it for the Board.1 Management presented the current inventory of AI systems in use2 and the proposal to expand the underwriting model.

    Directors confirmed the expansion falls within the acceptable-use policy and its human review thresholds3, and asked how a model failure would be escalated and, if material, disclosed.4

    Following the October education session5, directors questioned stressed-condition performance. The Board approved the expansion subject to remediation of two findings and a report back in March.6

    1. 1AI on the agenda, with a named owner
    2. 2A current inventory of AI in use
    3. 3An acceptable-use policy with human review thresholds
    4. 4Tested escalation and disclosure paths
    5. 5Documented director education
    6. 6A written record of questions asked and decisions made

    Illustrative minute, not a template. Build your record with Vela Boardroom →

    The evidence pyramid

    Principles are easy to state. The record is hard to produce.

    Most boards are strong at the base and thin at the top. Regulators, auditors and courts ask for the top.

    The recordWhat you can show you asked and decided
    TestingEvidence the controls work
    ControlsWho approves, monitors and can stop AI
    PolicyAcceptable use, risk appetite, escalation
    PrinciplesValues the board has agreed
    Hardest to evidenceEasiest to state

    Next 12 months

    Four moves that cover most of the landscape.

    Whatever your jurisdiction, these four leave evidence that answers most of the rules on this page.

    1Inventory

    List the AI in use, including what came inside vendor software.

    Evidence it leavesA current register the board has seen
    2Owner

    Name the committee and the executive who answer for AI.

    Evidence it leavesA charter line and a reporting cadence
    3Policy

    Agree acceptable use, review thresholds and when to escalate.

    Evidence it leavesA policy the board approved
    4Rehearsal

    Work through one AI incident before it happens.

    Evidence it leavesMinutes showing the questions asked

    Work through all four with Vela Boardroom →

    Where practice sits

    Most boards have the policy. Few have rehearsed.

    Where common AI oversight practices sit on the adoption curve today.

    1234567
    Innovators1Agent incident playbooks
    Early adopters2AI-specific D&O review3Rehearsed AI incident
    Early majority4AI inventory reported to the board
    Late majority5Named committee owner6AI in the risk register7Acceptable-use policy
    Laggards

    Based on Vela’s conversations with directors. Measured data will come from the Vela Barometer.

    Turn obligation into a defensible record.

    The Navigator translates these obligations into a governance baseline your board can act on, and evidence. Request a briefing to see how.