In February 2021, millions of Australians opened Facebook and found the news gone. The company was in a dispute with the government over a new law requiring platforms to pay for journalism. Overnight, it blocked news for everyone in the country. Some government health and emergency pages briefly went dark with it.
One month earlier, Twitter and Facebook had suspended the accounts of a sitting US president. Whatever one thinks of those decisions, the lesson was clear. A private company could flip a switch, and a nation’s public life would change.
Now imagine that switch does not control a news feed. Imagine it controls the system your hospital uses to read scans. It runs the fraud checks at your bank and the processing at your tax office. It powers the tools your children use to learn. This is not science fiction. It is where artificial intelligence is heading. That is why “AI sovereignty” has become one of the most debated ideas in global policy.
What AI sovereignty means
In simple terms, AI sovereignty is a country’s ability to make its own decisions about AI – how it is built, how it is used and how it is governed. Stanford University’s AI Index 2026 describes it as a state’s capacity to make independent decisions over AI’s development, deployment and governance. A coalition at the UN Internet Governance Forum adds a key point. A country must be able to understand and develop AI while keeping control and self-determination over it.
Sovereignty here is not a yes-or-no matter. It is a spectrum. The best comparison is energy security. No country produces everything it needs. The real question is whether someone else can switch off your lights.
We have seen this before
The worry is not new. In 1980, a UNESCO commission led by Seán MacBride published a report called Many Voices, One World. It found that global news flowed largely one way, from North to South. A handful of Western agencies decided what the world heard about Africa, Asia and Latin America. Developing countries called for a New World Information and Communication Order. The effort faded amid Cold War politics. But the concern was real, and it never went away.
Social media sharpened it. A few American platforms, and later China’s TikTok, came to shape what billions of people see, share and say. Their rules were written in California or Beijing, not in the countries where they applied. In Myanmar, the consequences were deadly. In 2018, a UN fact-finding mission found that Facebook had been used significantly in spreading hatred against the Rohingya. A foreign company’s failures in moderation fed a domestic human rights catastrophe.
Since then, switches have been flipped in every direction. Meta blocked news in Canada in 2023 over a law similar to Australia’s. The European Union banned the Russian broadcasters RT and Sputnik in 2022. The United States passed a law in 2024 to force TikTok’s sale or ban. Governments have also shut down the internet during unrest, from India to Iran to Bangladesh. Sovereignty can protect citizens. It can also silence them.
Private control now reaches even into war. Walter Isaacson’s 2023 biography of Elon Musk reported that Musk declined to extend Starlink satellite coverage for a Ukrainian military operation. One private individual made a decision with wartime consequences.
Why AI goes deeper
AI repeats all these patterns, but at a deeper level. Three differences stand out. First, social media carries your voice, while AI increasingly does the thinking, writing and deciding. Second, switching off social media silences people. Switching off AI could stall hospitals, banks and government offices that have come to rely on it. Third, a social media app can be copied cheaply. Advanced AI needs cutting-edge chips, vast amounts of energy and billions of dollars. Only a handful of players have all three.
That concentration is stark. A 2025 study for the Paris Peace Forum estimates that the United States controls about 75% of the world’s AI computing power. China holds about 15% and the European Union about 5%. Almost everyone else is a customer. Georgetown University’s Center for Security and Emerging Technology warns that dependence on foreign companies means access can be conditioned, restricted or revoked as geopolitics shift. US export controls on advanced chips show this is already happening.
Hosting data centres is not the same as controlling them. Researchers at Oxford University’s Oxford Martin School found that Britain hosts about 4% of the world’s AI computing capacity. Yet it has sovereign control over only around 0.1%. A foreign company’s data centre on your soil still answers to the laws of its home country. Some critics argue that large “sovereign AI” investment deals can lock in dependency rather than end it.
Also Read: OTS Summit in Ankara: Advancing Uzbekistan’s Strategic Priorities
Private hands on the switch
There is a further twist. The switch is mostly not in government hands at all. According to Stanford’s AI Index, industry produced more than 90% of the notable AI models released in 2025. US private investment in AI reached $285.9 billion that year, about 23 times the private investment in China. Stanford also notes that private firms increasingly build what governments call national AI infrastructure.
Social media taught us that a company’s business model, built on attention and advertising, can shape public life. With AI, a company’s business model will shape public services. Citizens voted for neither.
Not independence, but a say
None of this means every country should try to build its own AI from scratch. Full independence is not realistic for almost anyone. Even the United States relies on Taiwan for its most advanced chips and on the Netherlands for the machines that make them. Countries that shut themselves off from AI would simply fall behind.
The sensible goal is managed dependence. Countries must decide which dependencies they can accept and which they cannot. They should spread their reliance across several suppliers, so that no single company or country holds the only switch. And they must keep firm control of the most critical pieces – the data of their citizens, the systems that run essential services, and the rules that govern both.
We will all depend on others for AI. That much is certain. What remains open is whether we will have any say over the terms, and any hand on the switch.
This is the first of a four-part series on AI sovereignty.
The views presented in this article are the author’s own and do not necessarily reflect the views of Global News Pakistan
Political and International Affairs Analyst; Research Fellow, Bangladesh Neeti Gobeshona Kendra (Policy Research Center, NGK)
