Central Asia's Digital Sovereignty Dilemma: Homegrown AI Tools vs Global Platforms

Central Asia's Digital Sovereignty Dilemma: Homegrown AI Tools vs Global Platforms

The government workers in Astana don't have much choice anymore. By September 2025, Kazakhstan mandated Aitu, a homegrown messenger, across all government agencies, quasi-public organisations, and the Armed Forces. It was framed as a security imperative. President Kassym-Jomart Tokayev declared that "Kazakhstan has developed a domestic messenger, Aitu, capable of providing the necessary level of security."

Across the border, Uzbekistan was building its own digital fortress. MyID, the country's digital identity platform, reached 14.5 million users. OneID auto-registration and mandatory biometric SIM verification became the new normal. The message was clear: Central Asia wanted to control its digital infrastructure.

This isn't happening in isolation. Russia launched MAX as a mandatory app on devices in September 2025. China has been consolidating state control over AI for years. South Korea is enforcing the AI Basic Act. The global trend is unmistakable: nations are choosing digital sovereignty over openness, and 80 million Central Asians are caught in the middle of this geopolitical pivot.

But here's the tension nobody wants to discuss openly: the region is betting big on self-reliance whilst simultaneously needing global talent and technology to make it work.

The Aitu Play: Domestic Control, Uncertain Outcomes

BTS Digital and Kazakhtelecom built Aitu from the ground up. The messenger isn't just an app; it's a political statement. One million downloads in a country of 20 million people is a decent start, but the mandate is what matters. Government workers now have no alternative. Private citizens are watching.

The app comes with roughly 20 AI tools built in. That's impressive for a domestic product. But the uncomfortable question is: how many of those tools actually work well? How many are competing with proven alternatives that people already use? Aitu works, but it's not obviously better than the alternatives. It just has the backing of the state.

This mirrors what happened with Russia's MAX app. Mandatory adoption doesn't necessarily mean adoption because people prefer it. It means adoption because people have no choice.

In the AI industry, the most important resource is talent.

That observation should haunt policymakers. Because talent flows toward opportunity, and opportunity flows toward open platforms with global reach. If you're a gifted AI researcher in Almaty, are you going to build the next frontier model for Aitu, or are you going to move to San Francisco or Singapore where the infrastructure, funding, and user base are exponentially larger?

Uzbekistan's OneID Ambition: Paperwork to Platforms

Uzbekistan is taking a different approach. Instead of replacing messaging apps, it's consolidating identity and governance into a single digital spine. OneID auto-registration removes friction. Mandatory biometric SIM verification makes it real. The vision is "from paperwork to platforms" in a single generation.

This is genuinely clever policymaking. Digital identity is foundational. If you control identity verification, you control access to every service downstream. Uzbekistan is building the railway, not just individual stations.

But there's a cost. Data concentration creates surveillance potential. One breach, one malicious actor, one regime change, and 14.5 million citizens' biometric data is at risk.

Kazakhstan's Broader AI Bet: From Models to Supercomputers

Kazakhstan isn't just mandating Aitu. The country is building AI infrastructure from scratch. KazLLM, a multilingual AI model tuned for the Kazakh language, represents what policymakers call a "Cambrian explosion" of generative AI. Alem.ai, powered by an NVIDIA H200 supercomputer, ranks 86th on the global TOP500 list.

In January 2025, Kazakhstan's Law on AI No. 230-VIII entered force. A Digital Code is in development. The eGov SuperApp signals a preference for domestic control across every transaction.

This is infrastructure thinking, not just app thinking. But it's also expensive and requires sustained talent, funding, and partnerships. The UNDP and Kazakhstan recently launched a programme for open-source solutions, which suggests even policymakers recognise that going entirely domestic isn't viable.

Kazakhstan has developed a domestic messenger, Aitu, capable of providing the necessary level of security.

The Surveillance Question Nobody Wants to Ask

Critics aren't shy about the elephant in the room. Domestic control can mean better security. It can also mean surveillance. When government mandates a messaging app, it can monitor traffic. When biometric data is concentrated in a national cloud, it can be misused.

Uzbekistan and Kazakhstan haven't been flagged for egregious digital rights abuses compared to some neighbours. But the architecture is being built before the guardrails are. That's the wrong order.

The concerns are concrete:

• Neither country has data protection legislation comparable to the EU's GDPR
• Government-mandated apps create single points of failure and potential censorship chokepoints
• Biometric data, once compromised, cannot be reset like a password
• No independent civilian oversight of these AI and biometric systems exists
• The line between "security" and "surveillance" is one policy change away from being crossed

The Talent Trap: Building Domestically While Losing Domestically

Here's the paradox: Central Asia's digital future depends on keeping talent at home, but the most talented engineers and AI researchers are leaving.

Kazakhstan exported roughly USD 1 billion in IT services in 2025. It employs 200,000 digital workers. Those are respectable numbers for a country of 20 million. But the growth trajectory is fragile. Every brilliant Kazakh researcher who moves to the Bay Area to work on frontier models is a loss. Every developer who chooses to build for global platforms instead of domestic ones is a vote of no confidence in the domestic ecosystem.

You can mandate Aitu. You can't mandate brilliance.

The Global Comparison: Everyone's Building Walls

This isn't unique to Central Asia. China has spent years consolidating AI governance. Singapore is investing in AI upskilling and free tools to build workforce readiness. The difference is in the philosophy.

Some countries are building infrastructure to compete globally. Others are building walls to protect domestically. Central Asia is attempting both at once. That's ambitious but risky.

Compare how Alibaba's Qwen super app grew to 300 million users through market demand rather than government decree. The Chinese model is highly controlled, yes, but the best domestic apps in China succeed because they are genuinely excellent, not merely mandated. Central Asia hasn't yet produced an app that people would choose over WhatsApp or Telegram on pure merit.

What This Means for 80 Million Citizens

For ordinary people in Kazakhstan and Uzbekistan, this plays out in small, daily ways. Government services are faster because they're digitised. But those services are also government-controlled. Private alternatives are gradually squeezed out. The app you use to message friends is the same app the government uses to message you.

This isn't necessarily malicious. It's pragmatic from a state perspective. Domestic control means less foreign leverage, clearer data residency, and the appearance of self-sufficiency. But it also means less competition, fewer alternatives, and a slower pace of innovation.

Frequently Asked Questions

Is Aitu better than WhatsApp or Telegram?

Aitu is comparable in basic functionality but lacks the global user base. It offers local hosting and government backing, which appeals to security-conscious state actors. For ordinary users, the switching cost is mandatory compliance, not superior features.

Why is Kazakhstan building its own AI supercomputer?

Data sovereignty and strategic autonomy. By hosting Alem.ai domestically, Kazakhstan avoids reliance on foreign cloud providers and ensures that frontier AI research stays within national borders. It's an expensive play, but a logical one for a country serious about digital independence.

What happens if Uzbekistan's MyID data is breached?

The regulatory framework for data breach notification and remediation isn't as robust as in the EU or North America. 14.5 million biometric records represent a massive vulnerability if security fails. Unlike passwords, biometric data cannot be changed.

Are Central Asian AI tools actually competitive globally?

Not yet. KazLLM and other regional models are functional for specific languages and use cases but lack the scale and investment of models from OpenAI, Anthropic, or Google. Domestic mandates create users, not innovation.

Is digital sovereignty possible without sacrificing openness?

Theoretically, yes. Practically, every country that's tried has struggled. Singapore invests heavily in AI upskilling and maintains openness. China invests in domestic capability and restricts global platforms. Central Asia hasn't found its balance yet.

Drop your take in the comments below.