When Power Shifts — Algorithms, Governance, and Who Controls the Future

The Productivity Shock: 10 Agents, 80 Data Scientists

The fourth ABQ Dialogues session — titled "The Crystal Ball" — opened with a piece of testimony that set the analytical tone for the entire evening. Ciprian Jichici, CTO of FoundationaLLM.ai with thirty years in computer science, described the shift that had occurred in his own organisation over the preceding twelve months. One year earlier, the bulk of his operational value was delivered by a team of more than 80 data scientists he led globally. At the time of the session, that work was being handled by ten AI agents he coordinates each morning, giving them notes, making observations, then letting them execute throughout the day. The productivity shift: a factor of twenty.

This is not a claim about a distant hypothetical. It is a practitioner's account of a change already complete. And its significance goes beyond the productivity ratio. "This is the first time in the history of our civilisation," Jichici emphasised, "when we are talking about replacing humans not with other humans, but with something else." Every previous wave of labour displacement — mechanisation, electrification, computing — replaced specific human capabilities with tools that still required human direction, human maintenance, and human coordination. The current transition is categorically different: AI systems can now take on the coordination function itself, managing subtasks, adapting to new information, and producing outputs that were previously the exclusive domain of trained professionals exercising independent judgement.

The implications of this categorical difference are not yet fully absorbed, even by the practitioners closest to it. Jichici was careful to clarify that AI is replacing repetitive and pattern-based work, not creative problem-definition or genuine discernment. But the boundary between what counts as "repetitive" and what counts as "judgement" is moving faster than most institutions have registered.

Democracy as a Competitive Disadvantage?

Prof. Dr. Silviu Rogobete, political philosopher and former diplomat, and Dr. Andrei Nuțaș, AI ethics researcher, raised the session's most uncomfortable structural observation: democratic societies face a measurable competitive disadvantage in the race for AI capability precisely because of their democratic features. Regulation, public deliberation, consumer protection requirements, environmental review, and data governance frameworks are not defects in democracy. They are the mechanisms through which democratic societies protect citizens and maintain accountability. But they impose time and resource costs that non-democratic actors do not bear.

This is not a new observation in technology policy, but the speed of AI development makes it newly acute. A state actor that can direct national resources into AI development without public accountability processes, environmental review, or labour protection concerns can move faster along certain dimensions. A private corporation that can fire its ethics board once the AI race becomes commercially important — as at least one major AI company demonstrably did — can make deployment decisions that a more accountable organisation could not.

The response to this structural disadvantage cannot be to abandon the democratic features that generate it. Prof. Rogobete argued that democratic regimes must unite while they still can, and must develop both defensive and offensive tools for naming and attributing actions by actors who use technology to undermine democratic governance. France has already blocked platform access for government employees and banned social media for minors under sixteen. Australia has taken similar steps. The pattern suggests that democratic societies are beginning to recognise the competitive dynamics and respond — but whether the response is fast enough and coordinated enough remains deeply uncertain.

Europe's Eight Assets — and One Weakness

The session's most practically useful inventory came from Ciprian Jichici's systematic account of Europe's position in the technology sovereignty race. He identified eight critical factors for technological sovereignty: energy, water, chips, computing capacity, AI models, capital, human resources, and policies. On each dimension, he assessed Europe's position honestly.

The case for European optimism is real. The ITER fusion project represents the most advanced effort to achieve positive-energy nuclear fusion. ASML holds a monopoly on the extreme ultraviolet lithography equipment required for advanced chip manufacturing — every cutting-edge chip fabricated anywhere in the world requires ASML's technology. Europe has substantial water reserves and, as Jichici put it with 110% conviction, "very many very smart people on this continent." The human capital asset is genuine and significant.

But policy is the Achilles' heel, and the contrast with American approaches is stark. Jichici illustrated with Finland's VTT institute: three years ago they announced a quantum computing initiative and EU financing. Recently, with evident regret, they admitted the race was lost before it started. In three years, they attracted €50 million of investment. Similar American startups attracted a combined $70 billion over the same period. DARPA commits tens of billions with explicit declarations that they are "all in" on specific technological bets. The EU invests a bit here, a bit there — spreading resources across multiple initiatives, multiple countries, and multiple political priorities in a way that optimises for political legitimacy rather than competitive impact.

Prof. Drăgan, Rector of Politehnica Timișoara, offered the local analogue: Romania is the most engineering-focused city in the country by graduate ratio, yet this advantage has eroded since 1990. Only 200,000 Romanians invest in stocks from a population of 19 million. Seeds exist — drone programmes, microsatellite masters programmes, a European Space Agency hub. But seeds require watering, and the political vision and investment mechanisms to do that watering remain underdeveloped.

Algorithmic Governance: Who Decides?

Dr. Nuțaș provided the session's sharpest account of how technological choices become political choices without being recognised as such. He cited Microsoft's 2017 facial recognition system: over 90% accuracy for white men, around 80% for white women, under 7% for dark-skinned women. The data was correctly collected. The testing was correctly performed. The system performed exactly as designed. But "correctly" relative to what baseline? The question of who defines correctness, who selects the training data, who sets the performance targets — these are not technical questions. They are political questions dressed in technical language.

When Europe debates digital sovereignty, Nuțaș argued, the discussion ultimately reaches down to this level: which AI systems will govern European education, healthcare, financial services, and public administration? The choice of which AI to use in a classroom is not a procurement decision. It is a decision about what values will shape the judgement of the next generation of citizens — whose assumptions about identity, truth, authority, and productivity will be embedded in the systems that form discernment at scale.

Jichici raised a temporal dimension that compounds the accountability problem. We are currently in what he called "a happy place": AI systems produce answers that humans can still verify. It is very possible that within a decade or two, those systems will produce answers that cannot be verified by human experts working in reasonable time. As AI systems increasingly learn from AI-generated content rather than human-generated content, the values embedded in early systems will compound through subsequent generations of training. The window for deliberate governance is now, and it is closing.

"We have never before replaced humans with something categorically different. Every previous wave of automation replaced human muscle or specific skills. This replaces human judgement."

Four Thinkers, One Urgent Question

The panel's power came from the combination of disciplines assembled. Prof. Dr. Silviu Rogobete brought the historical and philosophical depth of a political philosopher who has also navigated the practical machinery of diplomacy — his analysis of power transitions from the Concert of Europe to the present moment grounded the AI governance discussion in the longer arc of how human societies have managed transformative power shifts before. Dr. Andrei Nuțaș brought the ethics researcher's refusal to accept reassuring framings at face value, consistently pointing to where the accountability gaps lie and where "ethics washing" substitutes for genuine accountability. Ciprian Jichici brought the practitioner's credibility — his accounts of what is actually happening inside technology organisations had an authority that purely theoretical analysis cannot match. Prof. Dr. Florin Drăgan brought the institutional perspective of a university rector managing an engineering faculty in a mid-sized European city, where the gap between the global technological conversation and the local capacity to participate in it is visible and urgent every day.

The four did not agree on everything. The practical possibility of meaningful AI regulation — Nuțaș's advocacy for state oversight of AI training versus Jichici's assessment that intervention in proprietary models is technically utopian — remained genuinely unresolved. But they agreed on the urgency of the question: the systems that are being built now, with the values and assumptions that are being embedded now, will govern the daily lives of millions of people who have had no meaningful input into their design.

The Accountability Gap

The evening's conclusion was not reassuring, but it was honest. Democratic societies are legitimately slower than non-democratic ones in deploying powerful technologies, because democratic legitimacy requires deliberation, dissent, and the protection of those who would be harmed. This is a feature, not a bug. But it creates a structural lag that, if unaddressed, means that the systems governing democratic societies will increasingly be built by actors with no democratic accountability to the people those systems affect.

Nuțaș turned the mirror on the audience with the question that most directly implicates consumers: all the problems enumerated come from consumption. We consume OpenAI, Claude, American cloud services. By consuming, we finance them. Are we willing to give up our comfort and use perhaps inferior European solutions for the collective good? The political philosopher's answer was clear: these decisions are not being made because consumers, collectively, want them. The gap between individual incentives and collective outcomes is exactly the problem that democratic governance exists to solve — if it can act before the window closes.

The function of the session, as Prof. Rogobete observed in closing, is not to provide answers but to surface questions that, once asked, cannot be unasked. The questions raised about accountability without sovereignty, about truth in a world of infinite AI-generated realities, about what it means to replace human judgement with something categorically different — these are now on the table. Whether the institutions that are supposed to answer them can move fast enough is the most important open question in European technology policy today.