The Sophists Have Infrastructure Now
AI governance discourse tends to treat its ideological opposition as a matter of preference - people who value innovation over precaution, speed over safety, growth over regulation. This is a charitable reading. A more accurate reading identifies a coherent ideological programme that treats democratic oversight of technology not as a reasonable constraint to negotiate, but as a civilisational error to be corrected.
Understanding that programme is not optional for anyone working in AI governance. You cannot regulate what you have not diagnosed.
The Intellectual Architecture
The tech-right intellectual tradition that shapes current AI infrastructure politics draws from several overlapping sources, and its genealogy is worth tracing carefully - because the ideas explain the actions.
Curtis Yarvin's neoreactionary programme ("the Cathedral") provides the foundational critique: democratic institutions, universities, mainstream media, and regulatory agencies form a self-reinforcing system of progressive control that prevents genuine progress. Regulation, in this view, is not the product of legitimate democratic deliberation - it is a mechanism of Cathedral power. The correct response is not to engage with regulatory processes but to build systems that route around them.
Marc Andreessen's "Techno-Optimist Manifesto" (2023) provides the positive programme: technology is the solution to all human problems, technological development should be unconstrained, and anyone who proposes limits on technological development is a mortal enemy of human flourishing. The manifesto explicitly positions AI safety researchers and governance advocates as the enemy. Not as well-meaning colleagues with a different risk tolerance. The enemy.
Peter Thiel adds the political economy and the most explicit philosophical statement of the programme. In 2009, writing for the Cato Institute, he stated: "I no longer believe that freedom and democracy are compatible." This was not a rhetorical provocation. It was a considered position drawing from Carl Schmitt - the German jurist who argued that political life is fundamentally defined by the friend/enemy distinction, that the sovereign is whoever decides the exception, and that liberal democratic consensus-building is a civilisational weakness. Thiel's business philosophy follows the same logic: monopoly is not a market failure but a goal; competition is for losers; dominance is for founders. Applied to AI governance, this translates directly: regulatory frameworks are instruments of the enemy, and the correct response is to achieve a position from which regulation becomes irrelevant.
JD Vance - yes, the US vice-president - provides the translation layer. Vance did not arrive at these positions independently. He was mentored by Thiel, who contributed approximately $15 million to his 2022 Senate campaign through affiliated PACs - one of the largest single-candidate political investments in modern US Senate history. Vance has acknowledged Thiel as an important figure in his conversion to Catholicism, a fact that matters because the intellectual formation was deliberate and the political investment was strategic. Vance now occupies the second-highest office in the US federal government, and the conversion of libertarian tech-right philosophy into nationalist populism is complete: Silicon Valley infrastructure dominance is reframed as American greatness, and regulatory oversight is reframed as elite hostility to the working class.
The network extends further. David Sacks, PayPal co-founder and ideological ally, is now the White House AI and cryptocurrency czar. The alumni of one payment processing company - which subsequently generated or backed Facebook, LinkedIn, YouTube, Tesla, SpaceX, and Palantir, enterprises now collectively worth several trillion dollars - have dispersed into positions of infrastructure control, political power, and governmental authority. This is not coincidence. It is what a coherent network with shared intellectual commitments and substantial capital looks like when it operates with strategic patience over two decades.
An Old Playbook
There is a striking structural parallel that predates Silicon Valley by twenty-five centuries.
In fifth-century Athens, the sophists occupied a position similar to the one the tech-right occupies today: embedded in the educational and political establishment, well-resourced, intellectually confident, and completely uninterested in truth as a constraint on effectiveness. The sophists' programme was explicit - they taught not wisdom but the technique of winning; not virtue but the appearance of virtue; not universal good but the skills required to defeat opponents and acquire power. For the sophists, moral relativism was not a philosophical position - it was a competitive advantage. Laws and institutions existed to be instrumentalised by those clever enough to do so.
Socrates spent his life arguing the contrary position: that universal Good exists, that virtue is knowable through honest inquiry rather than persuasive technique, and that democracy's institutions - however imperfect - were worth engaging rather than routing around. He lost the argument with a 281-to-220 jury vote and a cup of hemlock. The sophists, for their part, continued to thrive. They were very good at thriving. It was what they taught.
The parallel is structural, not historical. The tech-right programme is sophistic in precisely the Socratic sense: technique replaces truth, winning replaces virtue, and institutional engagement is replaced by infrastructure capture. The rhetorical sophistication of the "techno-optimist" and "founder mode" discourse conceals a programme that is, at its core, about eliminating the constraints that democratic accountability imposes on the accumulation of private power. Socrates would have recognised the move. He would have had questions.
He would not have been invited to the demo day.
The Infrastructure Play
What makes this ideological programme consequential - rather than merely interesting to intellectual historians - is that it has been translated into physical infrastructure at a scale that governance frameworks were not designed to address.
In May 2026, SpaceX's S-1 filing revealed a compute agreement with Anthropic worth $1.25 billion per month for access to the COLOSSUS data centre infrastructure - a contract totalling nearly $45 billion through 2029. Elon Musk's SpaceX is the physical substrate on which Anthropic - the AI company most associated with safety and responsible development - runs its most computationally intensive work. The safety company and the infrastructure company share a data centre. They do not share a theory of governance.
Anduril, founded by Palmer Luckey and operating within the PayPal Mafia's orbit, is building military AI hardware with Meta - smart glasses designed for warfare.
Palantir, with a market capitalisation exceeding $300 billion, holds contracts with the US military, ICE, and various intelligence agencies for surveillance, targeting, and deportation tracking. CEO Alex Karp has acknowledged the product's function plainly: Palantir systems "are used, on occasion, to kill people." The company's co-founder has written extensively about his view that democracy and freedom are incompatible.
The compute, the chips, the cooling, the power - the physical foundation of frontier AI development - is increasingly concentrated in the hands of a network of actors who share not just business interests but an intellectual framework that treats governance as the enemy. This is not a conspiracy.
Conspiracies require secrecy.
This programme has a manifesto, a Cato Institute essay, a Senate campaign, and an S-1 filing. It is a structural fact about who controls the infrastructure. And infrastructure shapes what is possible, regardless of what policy intends.
The Palantir Inversion
A Techdirt piece in April 2026 noted that Palantir employees were finally asking questions about the ethics of the projects they worked on - a reference to the Mitchell & Webb sketch about soldiers who realise, partway through the war, that their side has skulls on their caps.
The skull question is overdue. But the more significant observation is structural: private companies with explicit ideological programmes are building the AI-enabled surveillance infrastructure that states then purchase and operate.
The governance question - who regulates AI systems built by ideologically motivated private actors and deployed by state agencies - does not have a clean answer under the EU AI Act. The AI Act regulates deployers and providers. It does not regulate the intellectual frameworks that shape what those systems are designed to do, which is rather like regulating the speed of the car while ignoring who drew the road map.
The EU as Alternative Architecture
The EU's AI governance framework was built on different assumptions: that democratic accountability, fundamental rights, and human oversight are features, not bugs; that AI development should be subject to the same legal constraints as other technologies that affect people's lives; that the public interest in AI development is not identical to the private interest of AI companies.
These assumptions are contested. The tech-right programme is an explicit argument that those assumptions are wrong and that acting on them will produce civilisational decline. The argument is being made not in academic journals but through infrastructure investment, regulatory lobbying, and the placement of ideological allies in positions of governmental authority.
For EU AI governance to succeed, it needs to understand this opposition as what it is - a coherent programme with material resources behind it - rather than treating it as mere corporate self-interest to be managed through negotiation.
You do not negotiate with someone who believes the table you are sitting at should not exist.
Practical Implications
The Clause notes that the movement's preferred self-description changes with the political weather; the content does not. For governance purposes, the content is what matters:
Regulatory capture through complexity: complex technical standards that only large, resource-rich actors can comply with serve concentrated incumbents and exclude smaller competitors and civil society. Regulatory complexity is not a governance failure - it is a governance weapon. The compliance moat is not an unintended side effect of well-meaning regulation. It is a feature that the programme exploits deliberately.
Infrastructure leverage: control over compute, data centres, and foundational model weights creates dependencies that regulatory frameworks struggle to address. A company can comply with every obligation in the AI Act while the infrastructure layer - which is not subject to the same obligations - is controlled by actors with contrary interests. Anthropic's technical documentation can be impeccable. SpaceX's data centre does not file a conformity assessment.
Mission drift through commercialisation: the OpenAI restructuring, the Anthropic-SpaceX deal, the xAI infrastructure integration - each represents a step toward the consolidation of AI development in ways that weaken the mission commitments that motivated early safety-oriented work. The court that could have tested whether those commitments are enforceable ran out of time before it reached the question. The legal gap matches the ideological programme. It was not designed to. It did not need to be.
Democratic override through deployment speed: systems are deployed before governance frameworks are finalised, creating facts on the ground that regulators must accommodate rather than prevent. The AI Act's phased implementation timeline was anticipated by actors who front-loaded their most significant deployments. Malta's universal ChatGPT rollout is a case study: the deployment is scheduled to happen; the institutional response was silence; the governance framework moved from "constraining" to "accommodating" without anyone announcing the transition.
What Europe Can Actually Do
The response to an ideological programme is not primarily technical. It is political, institutional, and legal. But the question of what Europe can actually do - given that the other side controls the infrastructure and the most significant models - deserves a direct answer rather than an institutional aspiration.
Europe cannot win an infrastructure war against companies with trillion-dollar market capitalisations. It can win a governance war by a different mechanism: market access leverage. The EU single market is the largest in the world. Any company that wants access to EU consumers, EU public sector contracts, and EU data must comply with EU rules - regardless of where its infrastructure sits. This is how GDPR became a de facto global standard: not because EU institutions outbuilt Silicon Valley, but because the alternative to EU compliance was exclusion from a market that no serious company can afford to abandon.
The AI Act, the Digital Markets Act, and the Cyber Resilience Act extend this logic. The infrastructure is in Virginia and Texas. The market access obligation is in Brussels. These are not symmetric instruments - but the asymmetry is not necessarily unfavourable to the regulatory side.
The complement to market access leverage is sovereign capability investment. Mistral, EuroHPC, and the proposed EU AI compute reserve are not about matching US infrastructure scale. They are about ensuring that European institutions have genuine alternatives - that the choice between "use the US infrastructure on US terms" and "have no AI capability" is not the only available option. Investment in European AI infrastructure is a strategic hedge, not a competition. It needs to be funded and committed to accordingly.
The third tool is acquisition screening. The Netherlands demonstrated in May 2026 that infrastructure sovereignty can be defended at the point of acquisition, using the EU FDI Screening Regulation to block a US purchase of a digital supplier on national security grounds. More EU member states should be using this instrument more systematically. Sovereignty is cheaper to defend at the point of sale than at the point of deployment.
The honest accounting of the situation is this: the sophists of Athens won their institutional battles. Socrates was executed. His opponents considered the matter settled.
However...
The ideas survived - because they were articulated clearly enough to outlast the institutions that tried to suppress them, and because later generations found them worth preserving.
EU AI governance faces a comparable challenge - not execution, but institutional capture and regulatory neutralisation. The answer is the same one Socrates offered: articulate a better account of what good looks like, make it in public, in accessible language, and with concrete examples. Not as a philosophical counter-programme, but as a practical demonstration that governance serves people rather than the reverse. And, where Socrates fell short, have the institutional machinery to enforce it.
He had arguments. Europe has a single market.
The skulls are visible. The question is whether democratic institutions are looking clearly enough, and moving fast enough, to matter.