Newsletters
The Atlantic Council recently brought together the United States' leading tech policy minds alongside the senators who help turn those minds into law. Then it published the framework as a roadmap for how America intends to lead the world on artificial intelligence.
The document is serious, the names attached to it are serious, and the conference convened to launch it was the kind of event Washington does better than anywhere else: high-wattage, choreographed, and precise about what it wants you to take home.
Frederick Kempe, the Atlantic Council's president and CEO, opened with the line that set the frame for everything that followed: we are, he said, only at the "first 1%" of what artificial intelligence will become.
Senators Todd Young and John Hickenlooper were on stage, describing something they called a "combined arms" approach—a Marine Corps logic applied to technology policy, where energy, compute, and innovation don't just complement each other but must move in perfect sync or the whole formation breaks.
The frame was explicitly military, the stakes explicitly civilizational: out-compete China, export American "tech diplomacy," ensure that the next century runs on infrastructure Washington helped build.
Somewhere in the middle of it, my mind drifted to Ankara.
Not because Türkiye competes with the United States at anything resembling the same scale or league—it doesn't—but because the 1% framing carries a correct kind of pressure. If we are genuinely this early, then the policy choices being made right now are not current events; they are foundational. And if that is true, it matters enormously that one country is building an open field while another is laying the first stones of what looks, from a distance, increasingly like a fortress.
The case for the EU-type "AI shield" is not cynical. But it is, I think, wrong. Real sovereignty in AI will not arrive through the full-force securitization of the field, however loudly some call for it. It will arrive through commercial capacity—the ability to build, to compete, and to accumulate enough economic weight, actually to have political manoeuvre options. A country that cannot produce competitive AI companies does not get to set the terms; it only gets to comply with them.
The American vision is a bet on upstream dominance. The Commission's roadmap does not dwell on prohibition. It dwells on infrastructure: permitting reform for data centers, permanent compute access through the National AI Research Resource, and a proposed safety center built around voluntary standards rather than binding mandates.
Enforcement runs through civil tort reform and independent verification organizations, not criminal exposure. The logic is not to slow the machine but to ensure its brakes work, then let it run.
The Commission also talks about a "legion of entrepreneurs" emerging from community colleges, and invokes the four-day work week as a plausible downstream consequence of AI-driven productivity. It is civilizational transformation treated as a logistics problem: get the compute, fix the permitting, trust the market, win the race.
A widely held view in Turkish policy circles frames the global AI landscape not as an open field but as a hostile one, where Western models, trained on Western assumptions and deployed at scale, are distorting Turkish culture and history.
The prescription that follows from that diagnosis is state-led intervention: defensive, protective, aimed at neutralizing what proponents call the "poisoning" of national datasets by foreign systems.
Three separate AI legislative proposals are currently moving through the Turkish Parliament, each approaching the problem from a different angle. One aspires to a standalone AI statute modeled on the EU AI Act's risk-based architecture, but has eight articles against Brussels' 113; it is a sketch of a framework rather than the framework itself.
A second reaches directly for the Penal Code—introducing criminal liability for developers whose systems facilitate illegal acts, and a takedown window for certain illegal activities generated through AI.
A third is, in practice, a content-moderation bill: platform labeling requirements for AI-generated material dressed up as AI regulation. Together, they reveal that the legislature has recognized AI as a problem without yet agreeing on what kind of problem it is.
Parallel to the legislation, the institutional architecture is being assembled. Presidential decrees late last year created two new directorates—one for national AI inside the Ministry of Industry, one for public AI inside the Cybersecurity Presidency. The Medium-Term Programme 2026-2028 sets full alignment with the EU AI Act as a target.
Understanding why Türkiye is building a fortress requires understanding what the fortress is protecting.
"Digital Sovereignty" runs through every Turkish AI strategy document like a load-bearing beam. The goal is not simply to regulate AI—it is to ensure that Turkish language, Turkish culture, and Turkish data remain under national jurisdiction.
The 12th Development Plan identifies domestic large language models as a strategic priority precisely because dependence on foreign models trained on foreign corpora is framed in Ankara as a form of cultural exposure. Whether or not one accepts the framing, it has a real constituency.
The EU alignment logic runs alongside and reinforces this. Türkiye's candidacy relationship with Brussels has always produced a gravitational pull toward European standards, and AI is no different.
A Turkish tech company selling into the European single market needs to demonstrate compliance with Brussels' risk architecture regardless of what Ankara legislates.
Domestically aligning with the EU AI Act is less about pure ideology than commercial pragmatism wearing sovereignty language—a combination Ankara has always worn comfortably.
Congressional co-Chair Young's argument for the American model is essentially that voluntary standards preserve the faster lane, that prescriptive regulation extinguishes the innovation spark before it catches. If correct, that argument applies to Istanbul as directly as it applies to Austin.
A Turkish entrepreneur building an AI product operates under data protection obligations from the Turkish Data Protection Authority (KVKK), faces a prospective high-risk system registry, and navigates content regulation from the Information and Communication Technologies Authority (BTK).
Now, they potentially face a legal environment in which directing an AI system to commit a crime makes the developer its perpetrator.
None of that compliance surface exists in the American framework as currently designed.
The choice Türkiye is making is not irrational. EU alignment opens European markets. Criminal liability signals seriousness to a public that is reasonably anxious about algorithmic manipulation, or say, deepfakes. Sovereignty framing builds domestic political support.
But the Commission roadmap out of Washington is pointing toward a different kind of prize: the countries that let their entrepreneurs run at the speed of the technology rather than the speed of the compliance calendar will capture the upstream value chain.
Türkiye's current trajectory ensures it can trade with Europe. Whether it can compete with it is a different question, and one Ankara has not yet fully confronted.
If the 1% figure is right, the regulatory decisions of this year compound across decades. The American bet is that voluntary standards and market incentives can produce safety without sacrificing speed.
The Turkish bet is that a structured, EU-aligned framework can deliver both protection and participation, digital sovereignty and commercial competitiveness as partners rather than a trade-off.
Both bets are defensible, though neither is guaranteed. What the remaining 99% actually turns on is less about who writes the best code and more about whose governance architecture lets its entrepreneurs capture the gains before the rules harden into friction.
When it comes to geopolitics, Türkiye already engages far more closely with the United States than with Europe. There is no obvious reason why technopolitics should be any different.
Set as preferred source
Set as preferred source
8 min read
