Author: Sakina Arsiwala, a16z researcher; Source: a16z crypto; Compiler: Shaw Golden Finance

Lectures on YouTube: Content is a geopolitical weapon

Years ago, I was Google’s product lead for international search and subsequently led YouTube’s international expansion, launching the product into 21 countries in just 14 months. What I do is not only product localization, but also building local content partnerships and finding a way out of the minefields of laws, policies and market access. Most recently, I was responsible for community health (trust and safety) management for Twitch. I have also founded two startups during my career.

Today’s field of artificial intelligence (AI) bears striking similarities to the early growth stages of Google and YouTube. My career has taught me that globalization is not a product feature but a geopolitical game. The most profound lesson is that channel promotion is never purely a technical issue. Growth relies on local partners, cultural communicators, and trusted community opinion leaders who build bridges between global platforms and local users.

I personally experienced the GEMA copyright ban in Germany: a music rights agency almost excluded the entire country from YouTube's pan-European promotion plan. I have personally experienced the lese majeste arrest warrant turmoil in Thailand: As the head of external relations for YouTube, I faced the risk of being arrested for content on the platform that was deemed insulting to the King of Thailand, and could not even travel through the country. I have witnessed Pakistan shut down the internet across the country to block a video. I still remember our offices in India being physically attacked as global algorithms clashed with local religious taboos.

What we really need to deal with is never just a policy or infrastructure issue, but atrust barrier.

In every market, someone has to pay the cost to figure out what content is safe, acceptable, and valuable before users will participate. This cost will continue to accumulate, and over time it will form a kind of trust tax: it will be borne by a few groups in advance and then shared by everyone.

Today, the same contradiction is reappearing in the field of artificial intelligence, but the situation is more severe, evolving more rapidly, and the impact is more obvious. The recent standoff between the U.S. federal government and Anthropic has sparked public debate, while OpenAI has faced increasing scrutiny over its public sector partnerships. We are witnessing a shift: user acceptance is no longer solely determined by practicality, but increasingly ideological influence. In this environment, trust is very fragile, and a seemingly small collapse of trust may trigger a large-scale and rapid loss of users.

Google is doubling down on its deep trust strategy, leveraging user familiarity with existing Workspace and search ecosystems to gain access to the market, but the global landscape is becoming increasingly fragmented. The EU's strict regulatory red lines, China's fierce AI development competition, and growing AI nationalism have put the world on high alert.

The lessons of 2026 are clear: Institutional trust and cultural buy-in are now inseparable from the products themselves. It is impossible to build an intelligent operating system without trust as its cornerstone.

This is the barrier of sovereignty—the structural boundary where global artificial intelligence collides with local control. From a product perspective, it takes a more direct form: a trust barrier.

The expansion of all global artificial intelligence systems will eventually hit this wall. At this critical point, user acceptance no longer depends on technical capabilities, but on whether users, institutions and governments trust it in their own context.

The Internet used to be borderless. Artificial intelligence does not.

The end of the age of explorers

The first billion users of artificial intelligence are explorers and technological optimists. But the age of explorers has come to an end. In the past three years, we have lived in the era of prompt word engineering and digital alchemy. When people open popular applications such as ChatGPT and Claude, it is like going to a digital temple to witness the miracle of generative intelligence with their own eyes. In this day and age, the only metric that matters is model capability benchmarking:Who topped the latest benchmark? Who has the largest number of parameters?

But as we head into 2026, the bonfire of the Explorer Age is extinguishing. We no longer build toys for the curious, but instead turn to intelligent operating systems—the invisible, ubiquitous underlying channels that power individual entrepreneurs in Sao Paulo, Brazil, and community medical workers in Jakarta, Indonesia.

These users are not explorers, but practical users. They don’t want to talk to the “ghost” in the machine, they just want a tool that can help them solve the obstacles in real life. This is the real chasm-crossing moment for the next billion users. It is also in this unexploited edge area that Silicon Valley's dream of a global API has hit the cruelest reality of this era:Sovereignty Barriers.

The core change is: The popularization of artificial intelligence is no longer mainly a question of model capabilities, but a question ofdisseminationand trust. Frontier Labs will continue to improve model performance, but the next billion users will arrive not because a model scores better in a benchmark test, but because AI reaches them through institutions, creators, and communities they already trust.

Reality in 2026: AI becomes a national infrastructure proposition

In 2026, the core challenge for the industry is no longer making models smarter, but getting models acquired permission. The barrier of sovereignty is the boundary where general intelligence and national identity meet. Looking around the world, this barrier has begun to take shape: data localization requirements, national AI computing power plans, and model projects led by governments in India, the United Arab Emirates, and Europe. What began as a cloud infrastructure policy is quickly evolving into a smart sovereignty policy. Under this framework, the country refuses to become a “data colony” and requires intelligent systems that serve its citizens to run in its own sovereign data warehouse, inherit local culture, and respect national borders.

When you see the CEOs of Google (Sundar Pichai), OpenAI (Sam Altman), Anthropic (Dario Amodei), and DeepMind (Demis Hassabis) appearing on the same stage with Indian Prime Minister Modi at the 2026 India Artificial Intelligence Impact Summit, what you see is the true manifestation of the barriers to sovereignty. The M.A.N.A.V. vision put forward by Prime Minister Modi (ethical system, accountable governance, national sovereignty, inclusive AI, trusted system) sends a clear signal: if cutting-edge laboratories try to compete directly with consumers, they will eventually be eliminated by regulation. And trust is the only currency across these borders.

The dilemma of weakening network effects and why it forces new strategies

Unlike social platforms, which increase value for all other users with each new user, the value of artificial intelligence is largely localized. The thousandth prompt word I send out doesn't directly make the system more valuable to you. Although the data flywheel can optimize models, the user experience is always personal, not social. AI is a personal tool and can be emotionally charged, but at its core it is a practical tool.

This creates a structural problem: AI cannot rely on the compounding social network effect that the previous generation of platforms relied on to rise. In the absence of a native social graph, the industry can only fall into a high-consumption cycle, constantly chasing early users, heavy players and technological elites. This strategy worked in the era of explorers, but it cannot reach the next two billion users on a large scale.

More importantly, this model will completely fail in the face of sovereignty barriers. Becausewhen network effects are weak, trust does not form spontaneously but must be introduced from outside.

Transformation: From network effects to trust effects

If artificial intelligence cannot rely on social network effects to promote popularization, it must rely on another force: a network of trust. This is a critical change:

From acquiring users to empowering intermediaries

YouTube is able to scale at scale by leveraging the existing network of human trust. The same must be true for AI. Rather than trying to build a direct relationship with billions of users, a winning strategy should be:

• Empower those who already have user relationships;

• Leverage the trust they have already built up;

• Distribute intelligence capabilities through these channels.

Why it matters

In a world shaped by sovereignty barriers:

• Restricted distribution channels;

• Directly facing user mode is vulnerable;

• Trust is local, not global.

Without strong network effects, artificial intelligence cannot achieve scale through brute force and must rely on trust to penetrate. Artificial intelligence has no network effect, it has a trust effect.

Solution: The age of intermediaries is coming

How did YouTube gain a foothold in the international market? It does not rely on a better player, nor does it simply localize the interface text. The key to success is to become the platform of choice for people who already have local trust. In every market, the starting point for user acceptance is not YouTube itself, but identity anchors—those individuals and communities that already have the right to speak culturally:

• Bollywood fan page compiles rare Shah Rukh Khan footage for Dubai expat community

• American anime fanatics build an in-depth content ecosystem that has not been covered by mainstream media

• Local comedians, teachers and mashup creators transform global content into culturally appropriate formats

These creators don’t just upload videos, they interpret the Internet for audiences, act as trust intermediaries, and build bridges between overseas platforms and local users. YouTube's success lies in becomingthe invisible infrastructure that supports these identity anchors.

The core logic that has been ignored: the direct-to-consumer model hits the sovereignty barrier

Today, most AI companies still adhere to a direct-to-consumer mindset: Build better models → Present them in a chat interface → Acquire users directly.

This model is effective in the short term, but it cannot last long. Because in high-friction markets, users will not adopt new technologies directly, but through trusted people.

YouTube expands globally not by persuading billions of users one by one, but by empowering those who already have their audience’s trust. This is what invisible infrastructure really means:You don’t own the user relationship, you power the user relationship. At the scale level, this model has a stronger moat.

From Chat to Agents: Empowering Trusted Intermediaries

This is the key to moving from chat interfaces to agents. Chat is a tool for individuals, while agents are leverage for intermediaries. If we apply the philosophy of Anthropic executive Ami Vora - "Build products for the most tired people", then in many markets, these people are exactly the Trust Converters:

• Educators who adapt overseas ideas

• Entrepreneurs dealing with local bureaucracy

• Community leaders dealing with information overload

The way to win is to solve the trust delay they face - the gap between global intelligence capabilities and local practical scenarios. This requires a practical intelligent support system:

• For Educators: Sora / GPT-5.2 reinvents the curriculum - replacing the American football analogy with cricket, retaining the core meaning while still being culturally relevant.

• For individual entrepreneurs: The agent can not only interpret Singapore tax forms, but can also complete and submit them through local APIs.

• For community leaders: Overlay contextual memory for WhatsApp - extract structured action items from 10,000 messages, retain effective information and maintain community norms.

The core of the feasible model: solving the trust delay in the last mile

To understand why this model can scale, you must understand trust latency. In many parts of the world, the bottleneck is not access to technology but the time, risk and uncertainty required to build trust. Popularization of technology does not rely on advertising, but on endorsements.

The mistake most AI companies make is trying to centralize trust tax payment through branding, distribution, or product polishing, but trust cannot be scaled in this way.

The fastest way is to outsource the trust tax to people who have already borne this cost - local creators, educators and operators. They have already tried and made mistakes for the audience to find out what works, what doesn’t work, and what is really important in the local scene, and they have taken the risk for the audience.

By empowering these trust intermediaries:

• User acquisition cost approaches zero: distribution relies on existing trust network;

• User lifetime value improvement: Practical functions are adapted to local needs rather than being generalized;

• Popularization is accelerating: Trust is inherited directly and does not need to be accumulated from scratch.

Enterprises will get a pay-free global sales team out of thin air, with credibility, efficiency and depth far exceeding any centralized promotion strategy. You’re no longer building a product for your users, butproviding leverage to someone your users already trust.

This is exactly the path of YouTube’s global expansion,and it is also the only way for artificial intelligence to cross the barriers of sovereignty.

Sovereign Data Warehouse: Geographical Moat

The ultimate destination of the technological optimism advocated by Marc Andreessen is not to fight regulation, but to productize regulation. Competing with China's DeepSeek and Kimi, victory does not depend on ignoring national borders, but on controlling the data warehouse.

What is a sovereign data warehouse? It is a preferred resident localized instance of the model, running within a country's Digital Public Infrastructure (DPI) system.

• Geo-moat: By giving countries like India and Brazil digital sovereignty over models, weights and data, we are fundamentally reversing the control landscape. Intelligent capabilities are no longer mediated by overseas platforms, but autonomously governed within national borders. This is not to directly "block" external opponents, but to significantly increase their impact costs, reduce external dependence, and reduce the risk exposure of being controlled, having data extracted, or suffering unilateral intervention.

• Identity Anchor: Deeply bind the model to local culture and legal reality to build an insurmountable moat for general artificial intelligence.

• Feedback Loop: Solving extremely local details like Malaysian tax licensing is not a distraction but a model accelerator. This provides the base model with cultural resilience, allowing it to remain at the top of the world in intelligence.

There is a real contradiction here. The vision of artificial intelligence is to achieve general intelligence, but the trend of sovereignty is driving the entire ecosystem toward fragmentation. If each country builds its own technology stack, we will face the risk of incompatible systems, uneven security standards, and duplication of resources. The challenge facing cutting-edge laboratories is not just to increase the scale of intelligence, but to design an architecture that can achieve localized control without weakening the synergy of global capabilities.

Three major structural changes in the intermediary era

1. Artificial intelligence distribution will enter the existing trust network

Artificial intelligence will not achieve scale through independent applications, but will be embedded in instant messaging platforms, creator workflows, education systems, and small and micro enterprise infrastructure - because trust has already been established in these scenarios. In the absence of strong network effects, distribution must rely on existing interpersonal networks.

2. National AI infrastructure will become standard

Governments of various countries will increasingly require key artificial intelligence systems to deploy localized models, build sovereign computing power, or undergo regulatory review, which will accelerate the implementation of sovereign data warehouse architectures.

3. The creator economy will shift to an intelligent economy

Creators no longer just produce content, they deploy agents to perform real tasks for their communities. These agents will become extensions of trusted individuals, inheriting their credibility and delivering intelligent capabilities through the trust network.

Of course, there is another possible future: the emergence of an absolutely dominant assistant, deeply embedded in operating systems, browsers and devices, directly establishing a connection between users and models, completely bypassing intermediaries. If it comes to fruition, the trust layer will be built directly into the assistant.

But historical experience points to a more diversified pattern. Even the most dominant platforms—from mobile operating systems to social networks—ultimately rely on ecosystems for growth. Intelligence may be universal, but trust is always local. No matter which architecture wins in the end, the core challenge will not change: the spread of AI is no longer primarily a question of models, but a question of distribution and trust.

Conclusion: Niche markets are truly global markets

The greatest fallacy of the age of explorers is the belief that intelligence is a standardized commodity—a single global API that works the same in a Manhattan boardroom as it does in a village in Karnataka. Sovereignty barriers reveal a crueler truth: Intelligence may be universal, but universality is not.

States and local agencies don’t want a black box external system; they want control, the ability to adapt to scenarios, and the right to shape intelligence within their own boundaries. What they want is not ready-made applications, but underlying channels - infrastructure, security systems and computing power so that their citizens can build them independently.

The growth logic in 2026 is no longer looking for a universal set of user experiences, but product flexibility - allowing intelligence to adapt to local scenarios, supervision and culture without losing core capabilities. If we continue to pursue global consumers directly, we will always remain an alien layer — fragile, fungible, and destined to repeat the shocks I experienced at YouTube.

But when we turn to empowering intermediaries, the model will completely change: from chat interfaces to intelligent agents, from persuading users to empowering trust intermediaries, from fighting regulation to transforming regulation into a moat.

Scaling artificial intelligence does not rely on models, but on trust.

The winner in the AI race will not be the company with the smartest models, but the company that best empowers local heroes — teachers, accountants, community leaders — to increase their capabilities tenfold. Because at the end of the day, intelligence is delivered in systems, and diffusion happens among people.