Pakistan has announced that it is building its own Urdu artificial intelligence model, called Buraq. The announcement has created a lot of excitement, and for good reason. Most of the world’s best-known AI tools were built first for English, and Urdu speakers often get a weaker version of the same tool, or no version at all. A model built with Urdu and other regional languages in mind sounds like a direct answer to that problem.
But excitement needs to be balanced with accuracy. At the time of writing, Buraq is a promise, not a product. There is no public website for it, no demo you can try, no app you can download, and no release date you can mark on a calendar. This article explains what was actually announced, what Buraq is supposed to do, what foundations the government says it will stand on, and — just as importantly — what has not been confirmed yet. If you only remember one line from this article, remember this: Buraq has been announced, but it has not shipped.
What Was Announced in Seoul
Federal Minister for IT and Telecom Shaza Fatima Khawaja announced Buraq at the Global AI and Digital Summit 2026 in Seoul. The summit ran from 19 to 22 October 2026. The announcement came during a World Bank panel on 21 October titled “Enabling Digital Government Service Delivery through DPI and AI”. DPI stands for digital public infrastructure, which is the basic digital systems a country builds so that services can work online — things like digital identity, payments, and ways for government systems to share data safely.
The announcement was reported by The Express Tribune on 22 October 2026 and by TechJuice on 23 October 2026. Those dates matter, because they show how fresh this news is. This is not an old project being re-announced with new details. It is a very recent announcement, and many of the questions a normal user would ask — when can I try it, what can it do, how good is it — simply do not have public answers yet.
That is normal for a national announcement of this kind, but it is also the reason this article is careful with its wording. When a government announces a technology project at an international summit, the announcement itself is news. The product, if and when it arrives, will be separate news. Mixing the two up is how readers end up believing they can use something that does not exist yet.
What Buraq Is Supposed to Be
Buraq is described as a large language model trained on Urdu and other regional languages. A large language model is the kind of AI system that can understand and generate human language. It can, in principle, answer questions, summarise text, draft messages, and help with reading and writing tasks. Those are general capabilities of this class of technology, not a list of features confirmed for Buraq itself. No confirmed feature list for Buraq has been published.
Alongside the language model, the announcement also covers procuring sovereign compute, so that Pakistani data is processed inside Pakistan. Sovereign compute, in simple terms, means computing power that a country controls within its own borders, rather than depending entirely on computers and data centres located in other countries. The idea is that sensitive data does not have to leave the country to be processed.
Buraq sits inside a wider programme called the National Artificial Intelligence and Emerging Technologies Development Programme, shortened to NAIEDP. According to Origin Pakistan, reporting on 22 October 2026 from an APP report dated 21 October, that programme has two parts: first, domestic AI compute capacity, and second, Buraq itself. In other words, the plan is not only to build a model, but also to build the computing capacity that a model like this needs to run.
This distinction is important. A language model without enough computing power is like a car without fuel. It may exist on paper, but it cannot do useful work at scale. By naming domestic compute as the first part of the programme, the announcement is saying that Pakistan wants to build both the model and the machine power behind it. Whether that computing power has actually been bought, installed, or switched on is a separate question, and it is addressed honestly later in this article.
The Four Foundations the Minister Cited
The minister did not present Buraq as a project starting from nothing. She pointed to four foundations that she said already exist, and she summed up the approach with a clear line: “AI is the last layer, not the first.”
The four foundations cited were:
- NADRA digital identity, with over 100 million Pakistanis verified.
- Raast, the State Bank’s instant payments system.
- WASL, the national data exchange.
- The National Data Governance Policy.
The logic of that list is easy to follow. Digital identity lets a system know who a person is. Instant payments let money move quickly and safely. A national data exchange lets different systems share information in an organised way. A data governance policy sets the rules for how data should be handled, protected, and shared. AI, in this view, sits on top of all of that. It is the last layer because it uses the systems below it, rather than replacing them.
This is a useful way to think about national AI projects in general. A chatbot on its own does not fix a government service. The service underneath it — the identity check, the payment, the record — has to work first. The minister’s line suggests that the government understands that order. But understanding the order is not the same as having every layer finished and working perfectly. One of those four foundations, the Data Governance Policy, is still in draft form, as explained in detail below. That does not cancel the announcement, but it does mean the foundation is not yet final.
What Buraq Is Not Yet
This is the most important section of this article, and it deserves to be read slowly. Many readers will search for Buraq expecting to find a tool they can open today. They will not find one, and they should not be misled into thinking they have missed a launch.
The following table separates what has been announced from what has not been made available. It is based only on what the reporting up to late October 2026 actually confirms.
| Item | Status at the time of writing |
|---|---|
| Public announcement of Buraq | Announced in Seoul on 21 October 2026 |
| Public website | Not available |
| Public demo | Not available |
| App to download | Not available |
| API for developers | Not available |
| Release date | Not announced |
| Model size | Not published |
| Model architecture | Not published |
| Training data details | Not published |
| Development team details | Not published |
| Independent benchmark results | Not published |
TechJuice, on 23 October 2026, states the “promised, not shipped” position explicitly, and Origin Pakistan agrees. That agreement between two separate reports strengthens the point. This is not a case of one outlet missing a launch. The launch has not happened.
Because of that, this article deliberately does not include any “how to use Buraq” steps. There are no steps to give. Any website, app, or social media post claiming to offer Buraq for download or sign-up today should be treated with great caution, because no official public release has been confirmed. If you come across such a claim, do not enter personal details, do not pay anything, and wait for an official announcement through the Ministry of IT and Telecom or another official government channel.
The Sovereign Compute Question
Sovereign compute is one of the strongest ideas in the Buraq announcement. Many Pakistanis are rightly concerned about where their data goes when they use an online service. If data is processed in another country, it is subject to that country’s rules and systems, and the user has little control over it. Processing data inside Pakistan, on computing power Pakistan controls, is presented as a way to keep that data at home.
However, Origin Pakistan notes an important gap: the minister’s remarks do not establish that compute has been purchased, installed, or made operational. In plain language, the plan to have sovereign compute has been described, but the working computers themselves have not been shown to be in place.
This is not a small detail. Buying and installing large-scale computing capacity takes time, money, and technical work. It involves hardware, power, cooling, security, and skilled people to run it. Until that capacity is actually working, the data-sovereignty benefit of Buraq remains a goal, not a present reality. Readers should therefore treat “sovereign compute” as part of the plan for Buraq, not as a finished system that is already protecting their data today.
None of this means the goal is unimportant. Data sovereignty is a real concern for governments all over the world, and many countries are trying to build more of their AI capacity at home. It simply means that the honest description of the current position is “planned”, not “done”.
The Money Claim: 1 Billion Dollars Through 2030
You may see a figure of 1 billion dollars attached to this story. It is important to understand exactly what that figure is, and what it is not.
The 1 billion dollars through 2030 figure appears in TechJuice only. It does not appear in the Tribune report or in the Origin Pakistan report. According to TechJuice, it covers AI, regulation, and public-sector projects, including welfare, health, and tax, plus a local AI cloud involving Telenor Pakistan and Data Vault Pakistan.
Because it comes from a single source, it should be treated as a single-source ministerial commitment, not as a verified budget line. A verified budget line would mean that the money has been formally allocated through a budget process, with documents a reader could check. A ministerial commitment, by contrast, is a statement of intent or a target announced by a minister. It may become a budget later, but it is not the same thing as money already set aside and approved.
This article therefore does not present 1 billion dollars as confirmed government spending on Buraq. If you read another article that states the figure as a fact without any qualification, you now know to ask a simple question: who reported it, and is there a second independent source or an official budget document behind it? In this case, the answer is that only TechJuice reported it, and no verified budget document has been presented in the reporting reviewed for this article.
Why Urdu Matters — and What the 0.03 Percent Figure Does and Does Not Prove
The case for an Urdu AI model is strong on its face. Urdu is widely spoken in Pakistan, and it is one of the country’s national languages. Yet most global AI models are trained mainly on English and a handful of other widely used languages on the internet. When a language has less training data available, the model often understands it less well, makes more mistakes, and produces less natural-sounding text. A model trained properly on Urdu and other regional languages could, in theory, serve Pakistani users better than a model that treats Urdu as an afterthought.
TechJuice cites a figure of 0.03 percent for Urdu in Common Crawl web content. Common Crawl is a large collection of web pages that is often used to train AI models, so the share of a language in that collection is sometimes used as a rough indicator of how much online text exists in that language. If Urdu really makes up only a tiny share of that collection, it would help explain why global models struggle with it.
But that figure must be handled with the same care as the money figure. It appears in TechJuice only, and it is unattributed beyond Common Crawl itself. That means it has not been independently verified in the reporting reviewed for this article. It should not be quoted as a settled, independently checked fact. It is a single-source figure, and readers should treat it as an indicator of the problem being described, not as a precise measurement they can rely on without further checking.
The broader point does not depend on that one number being exact. Even without it, the basic problem is clear: Urdu speakers deserve AI tools that understand their language well, and building a model with Urdu at its centre is a reasonable way to try to close that gap. The question is not whether the goal makes sense. The question is whether the model, when it arrives, will actually be good enough, and that question cannot be answered until there is a model to test.
The Data Governance Policy Is Still a Draft
Data governance sounds dull, but it is the rulebook that decides how personal and government data can be collected, used, shared, and protected. If Buraq is going to process Pakistani data inside Pakistan, the rules governing that data matter enormously. A strong, final, enforceable policy would give users real protection. A draft policy does not.
Here, the reporting reveals a conflict that readers need to know about. Origin Pakistan checked the Ministry of IT and Telecom’s register on 22 October 2026 and found the Data Governance Policy 2026 still labelled “Draft” and for consultation. The Pakistan Digital Authority, according to the same report, said consultation concluded on 5 August, but feedback still had to be incorporated.
In other words, the policy is not yet final law. It is still a draft, with feedback still to be worked in. Safeguards under that policy must therefore not be described as final or enforceable. They are proposed safeguards, in a document that is still being finished.
This matters for anyone thinking about what Buraq will mean for their privacy. A promise that data will be kept at home is only as strong as the rules and systems that enforce it. Until the Data Governance Policy is finalised, and until the computing capacity and the model itself are working, the privacy benefit of Buraq should be described as intended, not guaranteed.
The “First to Sign” Claim — Wording Matters
Another claim attached to the Seoul summit is that Pakistan was the first country to sign the Digital and AI Compact at the summit. This claim also needs careful wording, because the reporting does not use identical language throughout.
Pre-summit reporting described Pakistan as “set to become” the first country to sign. Post-summit, TechJuice on 23 October 2026 says Pakistan “became” the first. Those two wordings are not the same. “Set to become” is a prediction before an event. “Became” is a statement that the event happened. The difference may seem small, but for a claim about being first in the world, it matters.
Until this claim is verified against the World Bank or the Ministry of IT and Telecom directly, it should not be asserted in this article as a settled fact. The fair way to report it is to say that TechJuice reported Pakistan became the first country to sign, while earlier reporting had phrased it as something Pakistan was set to become, and that the claim awaits confirmation from the World Bank or the ministry. That is not doubting Pakistan’s achievement. It is simply refusing to present a wording difference as if it did not exist.
Readers who care about this claim can check it themselves by looking for an official statement from the World Bank or the ministry about the compact and its signatories. If such a statement confirms the claim, the wording can be updated to match it.
What This Means for Ordinary Pakistanis Right Now
So what should an ordinary reader in Pakistan take away from the Buraq announcement today? The honest answer has three parts.
First, it is good news that Urdu and regional languages are being taken seriously at a national level. A future model that understands Urdu well could help students, teachers, small business owners, government workers, and anyone who prefers to read and write in Urdu. It could make digital services feel less foreign and more useful.
Second, nothing changes for you today. You cannot use Buraq today, because it has not been released. Your daily AI tools — whether you use an international chatbot, a translation tool, or a writing assistant — remain the same tools you used before the announcement. Do not uninstall anything, do not pay for a supposed Buraq app, and do not share personal documents with any service claiming to be Buraq until an official release is confirmed.
Third, the real test is still to come. When Buraq is eventually shown to the public, the questions to ask will be simple and practical: Can ordinary people actually use it? Does it understand Urdu well, including everyday, informal Urdu, not just formal written Urdu? Does it handle other regional languages as promised? Is Pakistani data really processed inside Pakistan, and under what final rules? How does it compare with existing tools, based on independent testing rather than only the claims of its makers? Until those questions can be answered with evidence, Buraq remains an announcement with potential, not a proven tool.
What to Watch Next
If you want to follow this story without being misled, watch for specific, checkable developments rather than vague excitement. A genuine next step would be an official public demo, a public website, or a developer API. Another would be a confirmed release date from the ministry, rather than a date guessed by commentators. A third would be independent benchmark results — that is, test scores measured by people other than the team that built the model, so readers can judge quality for themselves. A fourth would be finalisation of the Data Governance Policy, so that the rules protecting user data are no longer in draft form. A fifth would be clear evidence that sovereign computing capacity is operational, not just planned.
Each of those steps, if it happens, will be news in its own right, and this site will be able to report it with the same care used here: what is confirmed, who confirmed it, and what is still unknown. Until then, the safest summary is the one this article started with. Pakistan has announced Buraq, its own Urdu AI model, as part of a wider national AI programme. The ambition is clear, the foundations have been named, and the product has not yet arrived.
According to The Express Tribune, Federal Minister for IT and Telecom Shaza Fatima Khawaja announced Buraq at the Global AI and Digital Summit 2026 in Seoul, describing it as a large language model trained on Urdu and other regional languages, with sovereign compute so Pakistani data is processed inside Pakistan.
FAQs
Is Buraq available to use right now?
No. At the time of writing, Buraq has been announced but not shipped. There is no public website, demo, app, API, or release date, and no “how to use Buraq” steps exist yet. Any service claiming to offer Buraq today should be treated with caution until an official release is confirmed.
What is Buraq supposed to do?
Buraq is described as a large language model trained on Urdu and other regional languages, built as part of Pakistan’s National Artificial Intelligence and Emerging Technologies Development Programme, alongside domestic AI compute capacity. No confirmed feature list, model size, architecture, training data, or independent benchmark results have been published yet.
Will Buraq keep Pakistani data inside Pakistan?
That is the stated goal of the sovereign compute part of the plan, but it is not yet proven. The minister’s remarks do not establish that compute has been purchased, installed, or made operational, and the National Data Governance Policy is still in draft form, so data safeguards should not be described as final or enforceable yet.
Has Pakistan committed 1 billion dollars to Buraq?
Not as a verified budget line. The 1 billion dollars through 2030 figure appears in TechJuice only, covering AI, regulation, and public-sector projects plus a local AI cloud, and it should be treated as a single-source ministerial commitment rather than confirmed, allocated spending until an official budget document or a second independent source confirms it.
Also read
NotebookLM (Gemini Notebook) Urdu Guide: Summaries and Audio Podcasts in Urdu — Confirmed by Google