FluxForce AML and Fraud Blog

Agentic AI in UAE Government: What the 50% Target Means

Written by Sahil Kataria | Nov 20, 2026, 6:50:14 AM

Listen To Our Podcast🎧

• 1:55
Agentic AI in UAE Government: What the 50% Target Means
Secure. Automate. – The FluxForce Podcast
Your browser does not support audio.

What is agentic AI in UAE government?

The UAE wants half its federal services and operations on agentic AI within two years. Someone still has to sign off.

Agentic AI in UAE government is now a dated, Cabinet-level programme. On 23 April 2026 the UAE announced a framework to move 50% of government sectors, services and operations to autonomous agentic AI models within two years, and on 18 May the Cabinet set out who in each ministry owns the change.

This guide gives Chief AI Officers the timeline of the three spring releases and what each one set, explains what the 50% figure does and doesn't measure, and shows where human sign-off sits in a programme that openly talks about agents acting on their own. Last updated 19 November 2026.

In This Article, You'll Learn
  • What the UAE government announced about agentic AI, and who is accountable for delivering it.
  • The timeline of the 23 April, 12 May and 18 May 2026 releases, and what each one set.
  • What has happened in the rollout since May, and what "50%" measures.
  • How to build defensible decisions into the first services you move.

Request the timeline walkthrough

Two of the three releases were Cabinet meetings and 12 May was a progress review, and the timeline above shows what each one set. Bring the first service your entity plans to move, and we'll walk through how its agent keeps sign-off visible, in shadow mode, with a person making the call.
Get the timeline of the 23 Apr, 12 May and 18 May 2026 Cabinet releases and what each set

The three spring releases and what each set

Two of the three releases were Cabinet meetings, and 12 May was a progress briefing. 23 April set the 50% target and its owners, and 18 May set the roles of every ministry and federal entity.

  1. 23 April, Cabinet meeting. 50% of government sectors, services and operations on agentic AI within two years, with leaders assessed on adoption.
  2. 12 May, progress review. Al Gergawi briefed the Prime Minister on the requirements for a 50% shift.
  3. 18 May, Cabinet meeting. A work team in each entity, headed by its minister or head of entity, owns at least 50% of federal services and operations.
"
Direct answer

Agentic AI in UAE government is a dated Cabinet programme. On 23 April 2026 the UAE set a target to move 50% of government sectors, services and operations to autonomous agentic AI models within two years, and on 18 May the Cabinet made each minister or head of entity own the change.

Agentic AI in UAE government: what was announced?

The Prime Minister, Sheikh Mohammed bin Rashid Al Maktoum, made the announcement at the UAE Cabinet meeting of 23 April 2026. The official release says the project "aims to transform 50% of government sectors, services, and operations to implement autonomous Agentic AI models within two years". In the same release he described AI models that can "run an independent series of actions without human intervention".

Three features matter to anyone running AI inside a federal entity. The deadline is personal, because ministers, directors general and entities will be assessed on their ability to adopt the change. Execution has a named owner, a taskforce chaired by Mohammad Al Gergawi, the Minister of Cabinet Affairs, under the oversight of Sheikh Mansour bin Zayed. The rollout is phased, "based on continuous performance and impact assessment".

The same Cabinet meeting approved a government services data sharing policy "based on the principle of collecting data once and using it securely across entities". An agent that has to ask a customer for a document another ministry already holds isn't saving much time.

The UAE isn't starting cold. In 2017 it appointed the world's first Minister of State for Artificial Intelligence and launched the UAE Artificial Intelligence Strategy 2031. Microsoft's Global AI Diffusion Report for the second quarter of 2026 puts the UAE top for AI use, at 73.3%, against 18.8% of the working-age population worldwide.

The timeline: what did the 23 April, 12 May and 18 May 2026 releases set?

These three releases are often grouped together as Cabinet decisions. We read all three on mediaoffice.ae, the Dubai Media Office site. Two were Cabinet meetings. The middle one, on 12 May, was a progress briefing to the Prime Minister by Al Gergawi.

  • The 23 April Cabinet decision set the target that the 18 May framework turned into roles. The 18 May framework gave the target to a work team in each entity. Each work team picks one service for the 90-day launch.
From a Cabinet target to one named team and one service in 90 days.
Date (2026) Release Type What it set
23 April Cabinet meeting, chaired by Sheikh Mohammed bin Rashid Cabinet decision The target of 50% of government sectors, services and operations on autonomous agentic AI models within two years. A taskforce chaired by Al Gergawi under Sheikh Mansour bin Zayed. Phased rollout based on performance and impact assessment. Leaders assessed on adoption. Data sharing policy approved.
28 April Council renamed Prime Minister's approval The Ministerial Development Council became the Ministerial Council for Artificial Intelligence and Development, as part of the new federal operating model.
12 May Progress review Briefing, not a Cabinet meeting Al Gergawi briefed the Prime Minister on progress. The team was identifying the requirements and mechanisms needed for a 50% shift towards AI-driven government operations.
13 May Ministerial Council meeting Council review The renamed council reviewed how ready ministries and federal entities were to adopt agentic AI.
18 May Cabinet meeting at Qasr Al Watan Cabinet decision A governance framework setting the roles of every ministry and federal entity. A work team in each entity headed by its minister or head of entity, charged with at least 50% of federal services and operations, with its own targets and indicators. The Phase One service list in four categories. Training for 80,000 federal employees.

The first figure shows how the decisions cascade. The 23 April Cabinet decision set the target that the 18 May framework turned into roles. The 18 May framework gave the target to a work team in each entity. Each work team picks one service for the 90-day launch.

What the 12 May release set

Nothing new was decided on 12 May, but one sentence is worth reading. The release describes the aim as moving 50% of processes and services into models "capable of executing tasks and making decisions independently". It's the strongest statement of agent autonomy in the three releases.

What the 18 May release set

This is the operational decision. Each entity forms a work team "headed by the respective Minister or Head of entity", and the teams must move "at least 50 per cent of the federal services and operations into Agentic AI systems and operation models". The Phase One list covers citizens' services, residents' services, business sector services and general public services. The same meeting approved drafting a federal law on AI in health that "will address health data governance, safety and quality standards, legal liability provisions".

UAE government agentic AI plan: who runs it, and how?

The UAE government agentic AI plan runs from the Cabinet, which sets the target and approves frameworks, down to a named person in each entity.

Body or role What the official record says
Sheikh Mansour bin Zayed Oversees implementation and chairs the Ministerial Council for Artificial Intelligence and Development.
Taskforce and executive committee Chaired by Al Gergawi, who set out five implementation tracks at the May retreat, each led by senior ministers and officials.
Artificial Intelligence and Data Authority Approved on 14 June 2026 as the single national body for data, AI and digital government, reporting to the Cabinet and chaired by Al Olama. Its functions include ensuring compliance across federal entities.
Entity work teams Headed by each minister or head of entity.
Chief AI Officers A Chief Executive Officer for Artificial Intelligence role created in ministries and federal entities, with the officers selected in 2024.

Sources are the 14 June release for the authority and a September 2024 release for the Chief AI Officers.

Al Gergawi's tracks put numbers on both halves of the target. One track aims to move 50% of institutional operations, with specialised teams for human resources, finance, procurement and internal audit. Another aims to move 50% of government services. A third sets a tenfold increase in the speed of complex decisions, without saying how speed will be measured.

Abu Dhabi runs its own programme. Its Government Digital Strategy 2025-2027 aims to make it "the world's first fully AI-native government by 2027", backed by AED13 billion, and every Abu Dhabi government entity now has a Chief Data and AI Officer.

UAE agentic AI government rollout: what has happened since May?

The UAE agentic AI government rollout moved quickly after the 18 May decision. This chronology uses official releases, and the one press report is labelled.

Date (2026) What happened
20 May National retreat with more than 400 ministers and officials. The first federal agents launched for procurement, tax auditing, customer happiness and technical support. Al Gergawi's ministry already runs around 140 agentic AI assistants on its GovAI platform.
10 June Workshop with more than 300 participants from 50 federal entities. Each entity must pick one service or operation and launch it on agentic AI within 90 days.
19 June The Ministry of Cabinet Affairs plans to move 75% of its operations, services and specialised functions, and the Presidential Court 75% of its services, above the 50% floor.
23 July Ajman completed the UAE's first agentic AI trade licence renewal through a "headless" government service, with "human oversight where required".
9 August Strategic track launched under the principle "human leads, AI enables", including task classification frameworks.
August The National Committee for Agentic AI issued a unified guide for designing agent-powered government experiences.
9 September Ministerial Council reviewed phase one and planned phase two, including a shared technical ecosystem called FedAI.
29 September DEWA moved 16 of its 21 services, 76%, to agentic AI in under 90 days.
14 October Al Olama said every deployed federal agent must plug into an Agent Control Tower, audited by agents and by humans.
18 November Dubai approved a City AI Agent initiative and launched a Government Resources AI Agents ecosystem of 93 agents for HR, finance and procurement, expected to work across 85 entities.

The first federal agents support staff work. The procurement agent "supports procurement teams and streamlines sourcing procedures". The official progress report of 30 October says some services already complete transactions while others are still in development, and that approvals, exception handling and "human oversight where required" remain essential.

UAE government agentic AI goal: what does "50%" measure?

The UAE government agentic AI goal is easier to quote than to measure.

The unit changes between releases. The April release says "sectors, services, and operations", the 18 May release says "at least 50 per cent of the federal services and operations", and Al Gergawi's tracks split it into 50% of services and 50% of institutional operations. Ask your own entity which list it's being measured against.

The deadline is "within two years" in every official release we read, though the government portal u.ae goes further and says the framework mandates agentic AI in 50% of government sectors, services and operations "by 2028". Treat late April 2028 as the working date until the Cabinet says otherwise.

The 50% is a floor, not a cap. Each entity sets its own targets, and two have set 75%. No official source we found gives a federal percentage of progress towards 50%. DEWA's 76% is a Dubai government figure for one authority, not a federal total.

The best explanation of the number comes from an interview. Gulf News reported Al Olama saying the team mapped all government services and found "around 70 per cent" had some tolerance for AI, while for "30 per cent the risk of making a mistake is actually higher than our tolerance, so we're not even looking at them". That's a press report of spoken remarks, and no published policy sets out the mapping, but it shows some services are out of scope by design.

Who signs off when a UAE government agent acts?

The official record answers this in pieces. The April and May releases describe agents that execute and decide on their own. The documents issued since then put people back at specific points.

  • Every deployed agent plugs into the Agent Control Tower. The Agent Control Tower is audited by agents and by humans. The agent asks for customer consent before it acts. The agent sends any decision that touches livelihood to the accountable official.
A central log, consent before action, and a person on anything that touches livelihood.
  • The UAE Charter for the Development and Use of AI, issued in July 2024, "emphasizes the irreplaceable value of human judgment and human oversight over AI" and encourages "modifying or removing systems that pose risks".
  • The August unified guide requires "obtaining customer consent before taking action" and telling customers which actions need their intervention.
  • The Government Media Office guide says "professional responsibility and final judgement must remain human", with human review and institutional approval before publication.
  • Al Olama said every deployed agent must connect to the Agent Control Tower, with "an audit that is done by agents and an audit done by humans", and that a decision affecting "human life and livelihood" is "a red line".
  • Gulf News reported him comparing agent permissions to bank "maker and checker" controls.

Put together, agents act inside a permission level, every action is logged centrally, and decisions that touch someone's livelihood stay with a person. The record doesn't yet say which permission level applies to which service, or who signs off each one. That's left to each entity's work team.

The second figure shows that model. Every deployed agent plugs into the Agent Control Tower. The Agent Control Tower is audited by agents and by humans. The agent asks for customer consent before it acts. The agent sends any decision that touches livelihood to the accountable official.

Why the data protection law doesn't settle it

Article 18 of Federal Decree-Law 45 of 2021 gives a person the right to object to decisions from automated processing and requires the controller to "include the human element in reviewing automated processing decisions" on request. Article 2, though, says the law doesn't apply to "Governmental entities which control or process Personal Data". For federal agents, the human review step comes from Cabinet frameworks, entity policy and the guides above, not from the PDPL.

Private firms are in a different position. The Central Bank's February 2026 guidance note lists "effective human oversight" among its principles for licensed financial institutions. In the DIFC, Regulation 10 has covered autonomous and semi-autonomous systems that process personal data since 1 September 2023, and allows them only for human-defined or human-approved purposes. A bank connecting to a federal agentic service stays under those rules even when the agency isn't. Our guide to AI agents in financial services covers where autonomy stops inside a regulated firm.

Defensible decisions: the test that matters

The general industry term for all of this is the agentic AI platform, the software that gives agents their tools, identity and limits and records what they do. Our term for what it has to produce is defensible decisions. A decision is defensible when someone who wasn't there can see what the agent saw, which rule applied, what it recommended and who signed off.

The OECD's working group on agentic AI in government puts it this way. Once AI can act, the questions become "what the system is allowed to do", "under whose authority, with what audit trail", and "how its actions can be stopped or reversed". It adds that governments "have a higher duty than the private sector to explain, justify and correct the actions they take". The Agent Control Tower is the UAE's answer to the audit trail part. The permission rules and the sign-off record are the parts each entity still has to write.

A control tower log shows what an agent did. A defensible decision also shows who signed off and why.

Who owns each part of sign-off?

A FluxForce recommendation for splitting sign-off inside one entity. Agents prepare, and a person makes the call on anything that touches livelihood.

ComponentResponsibilityBoundary
Minister or head of entityHeads the entity's agentic AI work team and owns its targets and assessment indicators.Doesn't delegate accountability for the transformation to a vendor.
Chief AI OfficerOwns agent mandates, risk bands and the sign-off record for each service moved.Doesn't approve individual citizen decisions.
Accountable officialMakes the call on medium and high risk cases and on anything touching livelihood, with the prepared file.Doesn't sign off without the evidence record.
Agent Control TowerCentral connection point and log for every deployed federal agent, audited by agents and humans (as stated by the AI minister).No published technical specification yet.
AI agentsGather information, run checks, prepare recommendations and take low-risk reversible steps where allowed.Don't decide anything that touches livelihood, status or money.

A person makes the call on every medium and high risk case and on anything touching livelihood, customers consent before an agent acts, every agent action is logged centrally, and any agent can be stopped at once with its kill switch.

Where does the target fit in the UAE National AI Strategy 2031?

The 50% target is the latest step in a strategy that's almost a decade old.

What is the UAE National AI Strategy 2031?

It's the federal plan announced in October 2017, which aims for the UAE to be one of the leaders in AI in the world by 2031, under the UAE Centennial 2071 vision. The OECD.AI policy record lists eight objectives, including "To adopt AI across government services to improve lives" and "To ensure strong governance and effective regulation".

How do other governments handle sign-off for AI agents?

We found no other government with a public 50% target and a two-year clock, but several have written down where a person must decide.

Government Instrument Where a person decides
United States OMB memo M-25-21 (April 2025) An impact assessment is required "before deploying any high-impact AI use case", and affected people get a timely human review and a chance to appeal.
United Kingdom AI Playbook (February 2025) Principle 4 asks for "meaningful human control at the right stages". Services that affect a person's legal status or rights "must only use AI to support decisions that must be made by a human decision maker".
Singapore Model AI Governance Framework for Agentic AI (January 2026) Defines "significant checkpoints at which human approval is required" and limits agents' autonomy and access to tools and data upfront. In a written answer to Parliament in May 2026, the ministry said organisations "should not allow high stakes or irreversible actions to take place without human review".
European Union AI Act, amended by Regulation (EU) 2026/1744 Article 14 says oversight must be "commensurate with the risks, level of autonomy and context of use". Under Article 27, public bodies must assess the impact on fundamental rights before deploying most high-risk systems. Regulation (EU) 2026/1744 makes those rules apply from 2 December 2027 for Annex III systems.
Canada Guide on the Use of Agentic AI (September 2026) Two agent principles, "bounded autonomy and recoverability". Every agent action is logged "in a system that the agent does not have the ability to change", and actions that send, publish, approve, spend or update records need a review step unless the impact is low and trivially reversible.

The UAE's red line on livelihoods mirrors the UK's legal-status rule and Singapore's high-stakes rule, and its control tower plays the role Canada gives to a log the agent can't change.

Gartner expects at least 80% of governments to deploy AI agents for routine decision-making by 2028, and 70% of agencies to require explainable AI and human-in-the-loop mechanisms for automated decisions affecting citizen services by 2029. In June 2025 it predicted that over 40% of agentic AI projects will be canceled by the end of 2027, citing "inadequate risk controls" among the reasons. In business, Deloitte found only 21% of companies planning agentic AI report a mature model for agent governance.

Agentic AI for UAE government: where should a Chief AI Officer start?

Agentic AI for UAE government starts with service selection, and the federal workshop already gave the criteria. Readiness was judged on "the clarity and documentation of procedures, availability of up-to-date data, level of automation and outputs". Start where procedures are written down and the data is current, in five steps.

  • Each service goes into a risk band before an agent touches it. In the low-risk band, the agent may act on reversible steps and an officer checks a sample. In the medium-risk band, the agent prepares the case and an officer makes the call. In the high-risk band, every case goes to an officer whatever the agent recommends.
Sort first: low risk is sampled, medium risk is decided by an officer, high risk always goes to a person.
  1. Sort services into risk bands before you pick one. Low risk means a wrong action is cheap and reversible. High risk touches someone's livelihood, status or money, and stays with a person.
  2. Write the agent mandate in plain language, covering what the agent may do, which systems it may touch and what it must hand to a person. Canada's review-step rule is a good default.
  3. Run in shadow mode first. The agent works live cases without acting, and staff compare its recommendations with their own decisions each week.
  4. Record sign-off as data. For every case, keep the inputs, the rule in force, the recommendation, the person who signed off and their reason. That's what turns a control tower log into a defensible decision.
  5. Test the stop button. Every agent needs a kill switch that a named person can use, and every stop should be recorded. The Charter's call for "modifying or removing systems that pose risks" only works if someone can do it on an ordinary afternoon.

The third figure shows the sorting step. Each service goes into a risk band before an agent touches it. In the low-risk band, the agent may act on reversible steps and an officer checks a sample. In the medium-risk band, the agent prepares the case and an officer makes the call. In the high-risk band, every case goes to an officer whatever the agent recommends.

Internal operations, the HR, finance, procurement and internal audit work in Al Gergawi's operations track, are a sensible first move. They're documented and mistakes are easier to reverse.

What does this look like in practice?

The scenario below follows one federal procurement team through its 90-day service, using the risk bands, shadow mode and sign-off record described above.

Illustrative scenario: a federal entity's procurement team

Illustrative scenario, not a customer result.

A mid-sized federal authority picks supplier onboarding as its 90-day service, and the Chief AI Officer owns the agent mandate.

  1. The team sorts onboarding steps into risk bands. Document checks and register lookups are low risk. Approving a supplier above a contract threshold is high risk.
  2. The procurement agent runs in shadow mode for five weeks, checking documents and preparing a recommendation while officers keep working as before. The team reviews the differences weekly.
  3. The procurement head then lets the agent request missing documents on its own, a reversible step, with a weekly sample check.
  4. Supplier approval stays with an officer, who sees the agent's prepared file and records the decision and reason. Internal audit then replays ten closed cases from the record alone.

How does FluxForce approach agentic AI in government and regulated work?

FluxForce is the platform your departments run AI agents on. The agents investigate and prepare the work, a person makes the call, and every step leaves evidence an examiner can replay. We call it an Agentic OS for Regulated Industries.

  • Take in passes the department's records to understand. Understand passes one assembled view of the case to weigh. Weigh passes the checked and scored recommendation to prove. Prove passes the evidence record to report and learn.
Take in, understand, weigh, prove, report and learn, with governance and data protection on every step.

The platform works in five public stages: take in, understand, weigh, prove, and report and learn, with governance controls and data protection on every step. The fourth figure shows them in order. Take in passes the department's records to understand. Understand passes one assembled view of the case to weigh. Weigh passes the checked and scored recommendation to prove. Prove passes the evidence record to report and learn. Your rules run first as hard checks, and every step goes into tamper-evident evidence storage joined into one case file. Each agent works inside a mandate you set, you decide how much it does on its own in each risk band, high risk always goes to a person, and every agent has a kill switch.

Financial crime and compliance is the first app line on the platform: AI agents that investigate AML, sanctions, fraud and KYC alerts and prepare the case. Your analyst makes the call, with evidence an examiner can replay. You decide how much each agent does on its own, and every one has a kill switch. You can see the AI agents on the platform today, and how Policy and Compliance Management turns a written policy into controls that trace back to the clause.

Which sources support this guide?

UAE Cabinet, 23 April 2026 (Dubai Media Office)

50% target within two years, taskforce, phased rollout and leaders assessed on adoption.

Official release, 23 April 2026.

2026-11-19

Progress review, 12 May 2026 (Dubai Media Office)

Progress briefing by Al Gergawi, and models capable of executing tasks and making decisions independently.

Official release. A briefing, not a Cabinet meeting.

2026-11-19

UAE Cabinet, 18 May 2026 (Dubai Media Office)

Entity work teams, at least 50% of federal services and operations, Phase One list and 80,000 employees trained.

Official release, 18 May 2026.

2026-11-19

WAM: Agent Control Tower and red line (14 October 2026)

Agent Control Tower audited by agents and humans, and a red line on decisions affecting livelihood.

Remarks by the Minister of State for AI reported by the official news agency. No published technical specification.

2026-11-19

Where does Policy and Compliance Management fit?

A regulated firm has to turn its written policy into controls, much as an entity work team turns a Cabinet framework into agent mandates. FluxForce Policy and Compliance Management reads the policy and pulls out the rules it contains, ready for your compliance team to review.

You decide which rules go live, because every proposed control waits for compliance approval. Each control links back to the policy clause it came from, so an auditor can trace any alert to the line of policy behind it. The agent also maps new circulars to the clauses they affect.

We'll map one of your policies against your current controls and show you the coverage and gaps before you commit to anything.

See how Policy and Compliance Management works

The solution page today is written for AML, sanctions and fraud policies. FluxForce has no UAE government deployments and holds no UAE government certifications. It fits best today for a regulated UAE firm, such as a bank or insurer, that has to keep its own agent decisions defensible.

How do you plan the first 90 days of one agentic service?

A FluxForce recommendation that fits the 90-day federal launch window and keeps sign-off visible from day one. It is not a UAE requirement.

  1. Days 1 to 20. Sort candidate services into risk bands, pick one low or medium risk service and write its agent mandate.
  2. Days 20 to 30. Give the agent its own identity and least-privilege access, connect it to the control tower log and test the kill switch.
  3. Days 30 to 65. Run shadow mode on live cases and compare the agent's recommendations with staff decisions weekly.
  4. Days 65 to 90. Widen autonomy for one reversible step at most, have internal audit replay a sample of closed cases, and report results to the work team.
MetricDefinitionDecision guardrail
Recommendation matchShare of shadow mode cases where the agent's recommendation matched the officer's decision.Widen autonomy only where differences are explained.
Replay rateShare of sampled cases internal audit could replay from the record without asking anyone.We'd count a service towards 50% only once its sampled cases replay.
Key takeaways
  • The 23 April 2026 Cabinet set the 50% target with a two-year clock, and leaders are assessed on adoption.
  • The 12 May release was a progress review. 18 May gave each entity a work team that owns the target.
  • The 50% is a floor, and no federal progress figure is published.
  • Every deployed federal agent plugs into an Agent Control Tower, and livelihood decisions stay with a person.

Conclusion

The 50% target is real, dated and personal to the people who run each entity. The hard part is showing, case by case, who decided what and why, in a programme whose launch language talks about agents acting "without human intervention".

The record since May points one way, with customer consent before action, final judgement kept human, a control tower audited by people and a red line on livelihoods. Turning that into permission levels, sign-off records and stop buttons inside each entity is the Chief AI Officer's job.

Our view is simple. Move the services where you can prove every decision first, and count those towards 50%. A target met with decisions nobody can defend will cost more to unwind than one met a few months later.

Request the timeline walkthrough

Two of the three releases were Cabinet meetings and 12 May was a progress review, and the timeline above shows what each one set. Bring the first service your entity plans to move, and we'll walk through how its agent keeps sign-off visible, in shadow mode, with a person making the call.
Get the timeline of the 23 Apr, 12 May and 18 May 2026 Cabinet releases and what each set

Frequently Asked Questions

The UAE National AI Strategy 2031 is the federal plan, announced in October 2017, that aims for the UAE to be one of the world's leaders in AI by 2031. OECD.AI lists its eight objectives, which include adopting AI across government services to improve lives and ensuring strong governance and effective regulation.

UAE AI strategy is the set of federal plans and rules that guide how the country builds and uses AI. Its main parts are the National AI Strategy 2031, the July 2024 AI Charter, Chief AI Officer roles in federal entities and the 2026 agentic AI framework.

It covers government sectors, services and operations. The 18 May 2026 Cabinet decision says "at least 50 per cent of the federal services and operations", and Al Gergawi's tracks split it into 50% of services and 50% of institutional operations. The figure is a floor, and two entities are aiming for 75%.

No. In the UAE programme, agents prepare and execute work inside set permissions, and people stay accountable. The Government Media Office guide says final judgement must remain human, and Al Olama has called decisions affecting human life and livelihood "a red line".

Not for federal entities. Article 2 of Federal Decree-Law 45 of 2021 says it doesn't apply to governmental entities that control or process personal data, so its Article 18 right to object to automated decisions doesn't bind them. Private firms remain under the PDPL and, for banks, the CBUAE's AI guidance.

About the author

Sahil Kataria

Founder and CEO of FluxForce

Sahil works on secure AI and financial technology for regulated industries. His engineering background spans identity verification, payment security and compliance automation. He has led teams across Africa, the United States, Europe and India.

At FluxForce.ai, his focus is on explainable AI and auditable financial workflows. He writes for banking and compliance teams about the practical decisions behind these systems: how to assess risk, keep controls visible and introduce automation without losing accountability.

View Sahil Kataria author profile →
Policy and Compliance Management

Primary solution. A written policy turned into controls that trace back to the clause.

AI agents

The AI agents on the platform today.

AI agents in finance: what they do and where they fail

Where autonomy stops inside a regulated firm.

United Arab Emirates

UAE financial crime rules for regulated firms.