Who decides who may access an AI model? On 12 June 2026 the answer to that question was clear: citizenship. From the printing press to the internet, human history is full of attempts to control knowledge. Now it is AI's turn.
On the morning of 12 June 2026, the world did not yet know. But a small decision taken that day revealed something much bigger. The US Department of Commerce sent a letter to Anthropic. Its gist was this: no one who was not a US citizen would be able to access the most advanced models, Claude Fable 5 and Mythos 5. Turkish, Indian, Brazilian, German, it made no difference. If you were a foreigner, the door was closed. The justification was ready: national security. Anthropic objected, saying the risk was exaggerated and reminding officials that the same information was already available through other models. But the directive took effect. Because the company could not verify in real time who was a US citizen, it shut the models down completely for the entire world.
Eighteen days later, on 30 June, the Department of Commerce announced that it was lifting the restrictions. From 1 July, Anthropic reopened access to Fable 5 worldwide. Mythos 5, however, is still allocated only to selected organisations within Project Glasswing. The government reversed its decision, but once again did not explain why. What makes me think is not the technical side of this decision, but how familiar it is.
Throughout human history there has always been a dark relationship between power and knowledge. As knowledge spreads, power disperses; as power disperses, control weakens. In every age, this cycle has troubled those in power.
But the real issue is not having power; it is losing it. Psychologists have long known this: the fear of loss is a far stronger motivation than the desire for gain. People make twice the effort to avoid losing what they have as they would to gain it. This is true of ordinary people, and of states. Whoever holds power comes to live for protecting it. Over time this fear turns into paranoia, and paranoia into every kind of justification. The harshest censorship and the most ruthless bans in history were born largely not of malicious plans but of the reflexes of frightened people.
When the Church opposed the printing press, it probably did genuinely want to protect the faith. But at the same time, and perhaps more deeply, it wanted to protect its own authority. Even the clergy of the day could not have told which of these two intentions was stronger.
The same thing is happening today. Some of those restricting AI may have legitimate concerns. But beneath those concerns lies something that is never said out loud: if everyone has equal access to this tool, today's balance of power becomes meaningless.
Was it illiterate peasants who most fiercely resisted the printing press? No. They could not read anyway. Those who resisted were the clerical class who controlled access to knowledge and used Latin as a monopoly. The printing press destroyed that monopoly. Were the people most suspicious of the internet those who knew nothing about computers? No. Those who resisted were the big media organisations that held the monopoly on distributing news. The internet shook that monopoly.
Nor are the people most anxious about AI those far removed from technology. On the contrary: it is every kind of professional who produces analysis, writes reports, packages knowledge and sells it as value, earning a career, status and money in return. Because AI is cutting out that intermediary role.
A journalist can do in hours what took a research team days. A patient can understand their condition in depth before seeing a doctor. An entrepreneur keeps the corporate consulting fee in their pocket. Economic models and status hierarchies built over decades are being shaken. Those being shaken panic, and that panic turns into policy.
This US decision signals a new strategic paradigm. For years Washington had managed semiconductor exports within this framework. Who could buy Nvidia chips, and how much computing power each country could have, had become a strategic matter. Now the same framework is being applied directly to models, that is, to tools for thinking.
The logic is simple, the consequences heavy: whoever can think faster and deeper wins. In the economy, in science, in defence, in diplomacy. Controlling AI, in this sense, means controlling part of humanity's collective capacity to think.
A model a Turkish researcher cannot access gives an American rival a concrete advantage. A tool closed to a Nigerian engineer opens the way for a counterpart in Silicon Valley. This inequality is no longer only economic or educational; it is a cognitive and algorithmic divide. If this divide does not close, humanity will split into two tiers: those who think with AI and those who cannot.
Looking at history, we see this: attempts to keep knowledge under control have always failed in the long run. The walls of Latin came down. Radio frequencies were overcome. Internet censorship keeps being breached. There is only one reason for this: the human desire to know and understand has always proved stronger than the mechanisms built by those who want to stop it.
The same process will work for AI too. Open-source models are spreading, alternative platforms are being built, decentralised structures are developing.
The Fable and Mythos episode confirmed this, if on a small scale. The restriction lasted eighteen days and was then lifted. It was a short window, but what it showed was clear: even such a harsh measure could not last.
But this long-term optimism does not ease today's pain. The researcher whose access is closed today is left alone with the problem they cannot solve today. The hope that we will be free in the future cannot make up for today's loss.
Much of the discourse around AI is shaped by the language of fear. It will take our jobs. It will manipulate us. It will become uncontrollable. Some of these concerns are real. Like every powerful tool, AI can do serious harm in the wrong hands.
But fear does not have to be the only frame. The human brain evolved over millions of years. Language allowed thoughts to pass from generation to generation. Writing carried them beyond time. The printing press multiplied them. The internet made geography meaningless. AI is doing something different: it is expanding our capacity to process knowledge, that is, the speed and depth of thought.
Not replacing the human, but evolving together with the human. The sources a single brain can read in a lifetime are limited. The languages it can learn are limited. The fields it can connect are limited. AI stretches these limits; it enlarges human thought and lets it rise above its roots, not on top of them.
AI may remain in the hands of a few states and companies. Access is shaped by nationality, economic power and alliances. The cognitive divide deepens. History repeats itself; the tools change, the structure stays the same.
Or AI is positioned as the common heritage of humanity. Access is regarded as a universal right. Technology serves solutions, not conflicts. A global brain power is brought to bear on the climate crisis, on health inequality, on the education gap.
This choice is not a technical decision. It is a moral and political one.
The man inside the jar can see his surroundings. But the thickness of the glass distorts his view. It is hard for his voice to get out. His room to move is narrow. And most importantly, he depends on whoever holds the lid.
Billions of people whose access to AI is restricted are inside such a jar today. They see that things are happening outside. But they have no say in what is happening, no place in decision-making, no access to the tools that will shape the future. This is not a fate to be accepted. But to change it, we must first see it.
The printing press arrived, the Church resisted, the Renaissance was born. The internet came, the powers that be tried to rein it in, the world changed. AI is spreading, states are taking restrictive measures, then those measures are withdrawn; the cycle is the same, only the tool changes. But during the pain of transition, how many people, how much knowledge, how much potential will be wasted? And who will pay the price? We must not stop asking that question.
This article was first published in the August 2026 issue of Ekonomi Dünya magazine, in the column Kavanozdaki Adam (The Man in the Jar).
