Fridays, 2.30 pm (Lisbon) · Artificial intelligence on NOW →Fridays, 2.30 pm · AI on NOW →
Miguel Sousa SoaresMSS
← Writing
LinkedIn · Essay · 8 May 2026

Europe Has an AI Act. America Has the Trial over Who Controls It.

The West’s cognitive infrastructure is privately owned. The dispute over it is American. The consequences will be global. When the Oakland jury reaches its decision, Americans will have debated among themselves who controls the West’s AI layer. The decision will be issued by a federal court in California. Appeals will be heard in American courts. Europe will not be in the room. It will remain at the Commission, in Parliament and in public consultations, writing rules for the use of a machine built by others. Regulation matters. It civilises use, limits abuse and creates rights. But it does not replace ownership of the machine. The history of technological power is rarely written only by those who regulate. It is written by those who build, finance, operate and contest the infrastructure. Others arrived with the machine. We arrived with a law.

Artificial intelligence has left the laboratories, crossed the data centres, entered companies, reached governments and now entered a courtroom. Not as a product, but as a public promise, private infrastructure and a contest for the West’s cognitive power.

Since 28 April, in Oakland, Elon Musk has been taking OpenAI, the company he co-founded in 2015, to court in a federal case involving Sam Altman, Microsoft and other defendants. Musk testified during the first days. Greg Brockman followed on 4 May. Sam Altman and Satya Nadella are among the expected witnesses. Beyond the litigation lies a larger question: who controls the AI layer on which companies, states and citizens will work, decide and compete? On the surface it is a feud between founders. Beneath it is a struggle over the West’s cognitive infrastructure.

The case can be read as a billionaire soap opera or as an exercise in finding the villain. Neither reading is useful. Musk may have interests of his own; he founded xAI to compete on the same ground he is now challenging in court. OpenAI may have genuine industrial justifications and built one of the world’s most widely used systems. Both reveal the same thing: artificial intelligence has ceased to be a public promise and become private infrastructure. Generative AI is no longer merely a product. It is a layer mediating language, research, programming, analysis, decision-making and intellectual production. Whoever controls that layer controls more than software. They control an increasing share of how the economy thinks.

This transformation was not sudden. It was gradual, justified by the same force: scale. In 2015, OpenAI was founded as a non-profit organisation, with one billion dollars pledged by Musk, Altman and others, and a public mission: to ensure that advanced AI would benefit all of humanity. In 2018, Musk left the board. In 2019, OpenAI created a capped-profit structure and drew closer to Microsoft, which invested one billion dollars and formed a strategic cloud partnership. In 2023, the partnership became multibillion-dollar. In October 2025, OpenAI completed its recapitalisation: the non-profit foundation retained formal control of OpenAI Group PBC, with 26% of the equity; Microsoft held around 27%; investors and employees held the remaining 47%. In April 2026, that transformation reached court, with Musk as plaintiff and the company he helped found among the defendants. In little more than a decade, OpenAI had moved from a public promise to private infrastructure contested in court.

This is not a story of betrayal. It is a story about scale. Training a frontier model is no longer software engineering. It has become infrastructure engineering. It requires leading-edge chips in volumes produced only by TSMC. It requires data centres and energy on a scale few countries can provide reliably. It requires hyperscale cloud, concentrated among the three major American hyperscalers. It requires long-term contracts, capital intensity and exclusive privileges. The governance OpenAI designed in 2015 was not built for this scale. The mission could be universal. The machine that made it possible was anything but. The move to capped-profit in 2019 and to a new corporate structure in 2025 were not ideological accidents. They were concessions to the physics of scale.

The OpenAI case is no exception. It is the pattern. In the new AI landscape, no frontier lab is just a lab. Each is the visible face of a larger infrastructure. Behind OpenAI is Microsoft. Behind Anthropic are Amazon and Google, with the company valued at $380 billion in a private funding round in February 2026. DeepMind is Alphabet. Behind xAI is Musk, who raised six billion dollars in a Series C round in December 2024. The contest is no longer between laboratories. It is between entire ecosystems. The West’s cognitive layer is being built, powered and contested in the United States: American models, American cloud, American capital, American courts. Europe is watching from the outside.

Europe’s reading of this trial is not a complaint. It is a diagnosis. Europe’s absence is not procedural. It is industrial. Europe is not outside the room through distraction. It is outside because the room was built on assets it does not control. It will live with the decision but does not take part in the contest. Europe has regulation. The AI Act has been in force since August 2024. But this week the European Union itself reached a political agreement to postpone some high-risk rules until 2027 and 2028. Over the same period, Microsoft, Google and xAI agreed to give the US government early access to their models for national-security testing. Some test the machine with the state. Others delay the law’s application.

Europe has world-class scientific research in Munich, Paris and Zurich. It has a frontier AI company, Mistral, valued at around fourteen billion dollars in a private funding round in September 2025. It has talent. What it lacks is comparable scale. It has no hyperscalers on the American scale. It does not control production volumes of frontier chips. It cannot mobilise private capital at the speed this race demands. For there to be a European version of this litigation, there would first have to be a European version of this infrastructure. There is not. Europe has regulators. It has fewer litigants because it has fewer assets to contest.

The West’s cognitive infrastructure is privately owned. The dispute over it is American. The consequences will be global. When the Oakland jury reaches its decision, Americans will have debated among themselves who controls the West’s AI layer. The decision will be issued by a federal court in California. Appeals will be heard in American courts. Europe will not be in the room. It will remain at the Commission, in Parliament and in public consultations, writing rules for the use of a machine built by others. Regulation matters. It civilises use, limits abuse and creates rights. But it does not replace ownership of the machine. The history of technological power is rarely written only by those who regulate. It is written by those who build, finance, operate and contest the infrastructure. Others arrived with the machine. We arrived with a law.

Methodological note

Primary and secondary sources used in this article:

Musk v. OpenAI trial
Reuters, “Elon Musk says OpenAI was his idea, before executives looted it” (28 April 2026).
Reuters, “On witness stand, Elon Musk accuses OpenAI’s lawyer of trying” (29 April 2026).
Reuters, “Musk testifies he did not read ‘fine print’ about OpenAI becoming for-profit” (30 April 2026).
Reuters, “Key takeaways from Musk’s testimony at OpenAI trial” (1 May 2026).
Reuters, “OpenAI co-founder discloses nearly $30 billion stake, financial ties” (4 May 2026).
New York Times, “Here Are the Key Players in the OpenAI Trial” (28 April 2026).
Justia Dockets, Musk v. Altman et al., case 4:24-cv-04722, U.S. District Court for the Northern District of California, Judge Yvonne Gonzalez Rogers.

OpenAI timeline and corporate structure
BBC News, “Tech giants pledge $1bn for ‘altruistic AI’ venture, OpenAI” (December 2015).
TechCrunch, “Microsoft invests $1 billion in OpenAI in new multiyear partnership” (July 2019).
TechCrunch, “Microsoft invests billions more dollars in OpenAI, extends partnership” (January 2023).
OpenAI, “Built to benefit everyone” (28 October 2025).
TechCrunch, “OpenAI completes its for-profit recapitalization” (28 October 2025).
CNBC, “OpenAI completes restructure, solidifying Microsoft as a major shareholder” (28 October 2025).
Microsoft Blogs, “The next chapter of the Microsoft-OpenAI partnership” (28 October 2025).

American concentration and national security
Anthropic, “Anthropic raises $30 billion in Series G funding at $380 billion post-money valuation” (12 February 2026).
New York Times, “Anthropic Is Valued at $380 Billion in New Funding Round” (12 February 2026).
Reuters, “Anthropic clinches $380 billion valuation after $30 billion” (12 February 2026).
TechCrunch, “Elon Musk’s xAI lands $6B in new cash to fuel AI ambitions” (24 December 2024).
Reuters, “Microsoft, xAI, Google will share AI models with US government for security reviews” (5 May 2026).
Synergy Research / Statista, cloud market-share data (AWS, Azure, Google Cloud).

Europe, the AI Act and Mistral
European Commission, AI Act (Regulation EU 2024/1689), in force since 1 August 2024.
Reuters, “EU countries, lawmakers strike provisional deal on watered-down AI rules” (7 May 2026).
Modulos AI, “EU AI Act Delayed: The Omnibus Deal Closed on 7 May 2026”.
Mistral AI, “Mistral AI raises 1.7B€ to accelerate technological progress with AI” (September 2025).
Orrick, “Orrick Advises Mistral AI in its €1.7B Series C Funding Round” (September 2025).
CNBC, “AI firm Mistral valued at $14 billion as ASML takes major stake” (9 September 2025).

Note. Capped-profit is a hybrid corporate structure in which investors’ returns are capped at a multiple of their invested capital; anything above that limit goes to the non-profit parent entity.

Translation of the text in the author’s archive. Original publication: A Europa Tem Uma Lei Sobre IA. A América Tem o Julgamento Sobre Quem a Controla. · LinkedIn · Essay · 8 May 2026.
Original bibliographic details (Portuguese)

Ensaio de Miguel Sousa Soares, publicado originalmente no LinkedIn a 8 de maio de 2026. Esta página é o arquivo pessoal do autor.

Miguel Sousa Soares is an economist and analyst of geoeconomics and artificial intelligence, with more than twenty years’ experience in strategy and corporate finance. He writes for Visão, Jornal de Notícias, Expresso and Correio da Manhã. He comments on artificial intelligence on NOW on Fridays.