Skip to content

AI governance, risk and regulatory readiness

Every AI system controlled and owned, before the rules bind.

Watercolour of King Khalid International Airport in Riyadh: the control tower rises over the terminals, each aircraft stands in its own bay, and one moves along the taxiway.

Rules are coming. Prepare for them now.

AI governance in the Kingdom is moving from voluntary principles to binding policy. We put the inventory, the controls and the AI office charter in place before the first system goes live, then review each system against the Go-Live gate, so compliance shapes the design instead of delaying it.

The strategic question

Can we prove to a regulator or our board that every AI system is classified, controlled and assigned to someone accountable?

The problem

The rules are tightening. Few AI estates are ready to be asked.

AI governance in the Kingdom is moving from voluntary principles to binding policy.

More context

Meanwhile AI arrives through vendors, business units and staff tools faster than anyone can list it. When a regulator, an auditor or the board asks what you run and who owns it, the honest answer is often incomplete.

  1. No complete inventory

    No single record of each system and the data it touches.

  2. Compliance arrives too late

    Privacy and security questions surface at go-live, when change costs most.

  3. Accountability is unclear

    Nobody decides which uses are acceptable and which should stop.

What we do

Five lines of work. One accountable team.

  1. AI system inventory and risk classification

    We find and record every AI system you run or buy, give each one a risk class and a named person accountable, and set the rule for adding new ones.

  2. Policy, decision rights and the AI office charter

    A policy your audit and legal teams can work with, clear decision rights, and a charter for the AI office that holds them.

  3. Regulatory alignment

    Each system mapped against PDPL, NCA ECC and CCC, SDAIA's AI Ethics Principles and NDMO data governance, with every gap given an owner.

  4. Readiness for registration and audit

    Preparation for the registration and audit duties in SDAIA's draft Responsible AI Policy, labelled as a draft until it is final.

  5. Assurance before go-live

    A review of each system against the Go-Live gate. Models that inform credit, fraud and eligibility decisions are governed and assured through this review. We do not build them.

The rules, one by one

SDAIA, NCA, PDPL and the National AI Index. What each one asks, and how we help.

Four sets of rules shape AI governance in the Kingdom. Each card gives the publisher, the status in November 2026, what it asks of an AI system, and the work we do on it. A practical guide to start from; your legal and security teams sign off.

  • SDAIA's AI Ethics Principles and AI Adoption Framework

    Saudi Data and AI Authority (SDAIA) · Ethics Principles published September 2023; Adoption Framework version 2, May 2025, guidance

    What it asks

    • Seven ethics principles for anyone who designs, develops, deploys or uses AI in the Kingdom, from fairness to accountability and responsibility.
    • A risk tier for every AI system: little or no risk, limited, high, or unacceptable. High-risk systems need conformity assessments before and after deployment.
    • Named people responsible for AI ethics in each adopting entity.
    • Seven components for adoption, from strategy and governance to applications and impact, with an AI office and KPIs such as the move from pilot to production.

    How we help

    We build the inventory of AI systems, give each a risk tier and a person accountable, draft the AI office charter, and map your initiatives to the framework's seven components.

  • NCA's Essential and Cloud Cybersecurity Controls (ECC and CCC)

    National Cybersecurity Authority (NCA) · In force, 2024 editions (ECC-2:2024 and CCC-2:2024)

    What it asks

    • The ECC apply to government agencies and their affiliated companies, and to private owners and operators of critical national infrastructure. NCA encourages every other entity to use them as good practice.
    • ECC-2:2024 has four main domains and 108 main controls, including third-party and cloud computing.
    • The CCC set controls for cloud service providers and for the entities that use cloud services.
    • A model provider, a hosting platform and an integration partner each count as a third party to assess.

    How we help

    We design hosting, access and logging against the controls your organisation is subject to, assess the AI suppliers as third parties, and document the evidence for your security team.

  • Personal Data Protection Law (PDPL)

    SDAIA (regulator) · In force since 14 September 2023; amendments to the Implementing Regulation in draft (consultation 6 October to 5 November 2026)

    What it asks

    • Process only the personal data the use case needs.
    • Transfer personal data outside the Kingdom only under the conditions and safeguards in the transfer regulation, with a risk assessment where required. Sending prompts with personal data to a model hosted abroad can count as a transfer.
    • Run an impact assessment before high-risk processing, such as linking datasets or decisions based on automated processing.
    • Notify SDAIA within 72 hours of a breach where notice is required, and answer data subjects' requests within 30 days.

    How we help

    We map the personal data in each AI workflow, decide the hosting and any transfer with your data protection team, and run the impact assessment during design, while its findings can still change the design.

  • The National AI Index

    SDAIA · Launched July 2025 for government entities; methodology version 1, May 2025

    What it asks

    • Three pillars (Directions, Enablers, Outputs), seven sub-pillars and 23 domains, scored into six maturity levels from Absence of Capabilities to Pioneer.
    • Supporting documents and evidence for every answer, reviewed and verified by a specialised team.
    • An AI unit with a mandate, compliance with SDAIA's guidelines, and impact measured with return on investment models.

    How we help

    We run a self-assessment on the 23 domains, build an evidence folder for each sub-pillar, and set a 90-day plan for the gaps that move the score most.

How an engagement runs

Four stages, applied to this practice. A gate after each.

Each gate is a decision: continue, change course or stop. That keeps spend tied to results, and keeps you in control.

The Live Path

  1. Discover

    We build the inventory and the gap map, and rank your systems by risk.

    Gate A ranked roadmap with business cases.

  2. Design

    Risk class, controls and data flows are decided with your teams before design sign-off.

    Gate Design sign-off and a value baseline.

  3. Lead the build

    Controls are tested on the real system, with evidence gathered for each one.

    Gate The Go-Live Gate Checklist, signed.

  4. Run & transfer

    Monitoring and periodic review keep each system inside its risk class, and the AI office holds the record.

    Gate Value signed off by finance, and the next operation chosen.

We count systems running in live operations, not pilots.

See The Live Path in full

What you get

What your team keeps.

What you get

  • AI system inventory and risk classification guide
  • AI policy, governance charter and decision rights
  • Regulatory alignment map (PDPL, NCA, SDAIA, NDMO)
  • Pre-go-live assurance review for each system

Typical outcomes

  • A clear answer when a regulator, an auditor or the board asks what AI you run.
  • Compliance questions answered in the first workshop, not after go-live.
  • A governance model your audit and legal teams can work with.

How we measure

What we baseline, then track.

Each measure gets a baseline in the Design stage and is reported against it once the system is live. This is how we measure, not a result we promise in advance.

  1. AI systems inventoried and risk-classed

    The share of the AI systems you run or buy that are in the inventory with a risk class.

  2. Systems with a named person accountable

    The share of inventoried systems with a named person accountable.

  3. Reviewed before go-live

    The share of new systems that passed the assurance review before they went live.

  4. Gaps closed on time

    Gaps closed against those found, by the dates in the remediation plan.

How we measure, step by step

Ways in

Start with a fixed scope. Then decide what comes next.

See every way in
  • 3 to 4 weeks · fixed scope

    AI governance and regulatory readiness review

    Know where each AI system stands against the rules, and what closes the gaps.

    • An inventory of AI systems, each with a risk class
    • A gap map: PDPL, NCA ECC and CCC, SDAIA's ethics principles, the draft Responsible AI Policy
    • A draft AI office charter
    What's included

    Know where every AI system you run stands against the Kingdom's rules, and what it takes to close the gaps before the draft policy becomes binding.

    Best for: Government and semi-government entities, and regulated companies.

    • A 90-day remediation plan
  • 4 weeks · fixed scope

    The 4-week diagnostic

    AI opportunities ranked by value, with business cases and a roadmap.

    • Interviews with your teams in Arabic and English
    • A map of processes, data and handoffs
    • Opportunities ranked by value, data, fit and risk
    What's included

    Four weeks, fixed scope. We map how the work really happens, then hand you AI opportunities ranked by value, with business cases and a roadmap. You decide what comes next.

    • Business cases for the top initiatives
    • A roadmap with gates for the next stages

Questions

Questions buyers ask.

What does the PDPL mean for an AI system in Saudi Arabia?

It shapes three decisions: which personal data to process, whether data may leave the Kingdom, and how fast to report a breach.

More detail

The Personal Data Protection Law (PDPL) came into force on 14 September 2023, and SDAIA gave organisations a year to bring their processing into line. Purpose: process only the personal data the use case needs. Transfers: personal data can leave the Kingdom only under specific conditions and safeguards, with a risk assessment, and calling a model hosted abroad can count as a transfer. Breaches: the implementing regulation requires notifying SDAIA within 72 hours of becoming aware of a breach that could cause harm. Octopus designs AI systems in line with the PDPL from the first workshop. This answer is a practical guide to start from; your legal and security teams sign off.

How can an organisation prepare for SDAIA's draft Responsible AI Policy?

Know which AI systems you run, class each one by risk, and map the gaps against the draft before it becomes binding.

More detail

The Responsible AI Policy is still a draft; its public consultation ran from April to May 2026. According to a November 2026 legal briefing, the draft would make AI governance binding: risk tiers, registration of AI systems, safety reports before high-risk deployment, and annual third-party audits for high-risk systems. The final text may change. Octopus's AI governance and regulatory readiness review takes 3 to 4 weeks with a fixed scope. It delivers an inventory of AI systems, each with a risk class; a gap map against PDPL, NCA ECC and CCC, SDAIA's AI Ethics Principles and the draft policy; a draft AI office charter; and a 90-day remediation plan.

Does Octopus certify AI systems against PDPL, NCA or SDAIA?

No. The publishers own these frameworks, and Octopus does not certify against them or claim a certificate from them.

More detail

Octopus designs and reviews AI systems so they are aligned with PDPL, NCA's ECC and CCC, and SDAIA's AI Ethics Principles. Your legal and cybersecurity teams have the final word.

Who can help us apply SDAIA's AI Adoption Framework and the PDPL in Saudi Arabia?

Octopus does, from Riyadh and Al Khobar, through the AI governance and regulatory readiness review: a fixed scope over 3 to 4 weeks.

More detail

We map each AI system to the framework's seven components and to PDPL, NCA and NDMO requirements, list the gaps, and name a person accountable for each one. We work with companies and government entities across Saudi Arabia and the GCC. The work is aligned with these frameworks, and your legal and security teams sign off.

What does SDAIA's AI Ethics Principles document ask of an organisation?

Seven principles, a risk tier for every AI system, and named people responsible for AI ethics.

More detail

The principles apply to anyone who designs, develops, deploys or uses AI in the Kingdom: fairness; privacy and security; humanity; social and environmental benefits; reliability and safety; transparency and explainability; and accountability and responsibility. Systems with little or no risk carry no restrictions, limited-risk systems follow the principles, high-risk systems need conformity assessments before and after deployment, and systems with unacceptable risk are not allowed.

Which NCA controls apply to an AI system?

Usually the third-party and cloud controls of the ECC, and the CCC where the system runs on a cloud.

More detail

The Essential Cybersecurity Controls (ECC-2:2024) apply to government agencies and their affiliated companies, and to private owners and operators of critical national infrastructure; NCA encourages every other entity to use them as good practice. A model provider, a hosting platform and an integration partner are all third parties. The Cloud Cybersecurity Controls (CCC-2:2024) cover cloud service providers and the entities that use them.

What is the National AI Index, and how do we prepare for it?

SDAIA's measure of how ready government entities are to adopt AI, built on evidence for every answer.

More detail

SDAIA launched it in July 2025. It scores an entity on three pillars (Directions, Enablers, Outputs), seven sub-pillars and 23 domains, and places it on one of six maturity levels. To prepare, name a person accountable for the response, score yourself on the 23 domains, and build an evidence folder for each sub-pillar before the cycle opens. Our guide to the National AI Index lists the evidence to keep.

What evidence should we keep for an audit, a regulator or the National AI Index?

An inventory of AI systems with a risk class and a person accountable for each, and the records behind every control.

More detail

Keep the AI policy and the decision that set up the AI office, the PDPL impact assessments and the record of processing, hosting and transfer decisions with their risk assessments, supplier assessments against NCA's controls, the pre-go-live review of each system, and measured results against a baseline. The same folder answers an auditor, a regulator and the National AI Index questionnaire.

Need to know where your AI systems stand before the rules become binding? Start with the governance review.

30 minutes with our team. We tell you plainly where we would start.