The EU AI Policy Pivot: Adaptation or Capitulation?

Brussels’ AI policy is not what it used to be. Just 11 months ago, when the European co-legislators adopted the AI Act, the EU boasted of its regulatory superpower status. Further rules were in the works: detailed guidelines—so-called Codes of Practice, secondary guidance from the Commission, technical standards, etc.—to help companies work towards legal compliance. An AI Liability Directive was in the pipeline, too, to help assign blame (and financial obligations) when AI systems harm people.

How different is the situation today? At the beginning of the year, enforcement of the Digital Services Act stalled. Earlier this month, the European Commission announced that it put the AI Liability Directive on ice, too. And EU Commissioner for Tech Sovereignty, Security and Democracy Henna Virkkunen announced that the AI Codes of Practice would mostly serve to “help and support” AI companies rather than restricting them.

What are we to make of this AI policy pivot? Is it a capitulation to Trump’s aggressive tech and ‘America First’ policies? Or is the regulatory loosening a sign of intentional strategy, the adoption of rules that adapt to a changing environment and that are better suited to boost domestic companies and EU policy goals? How we answer that question is not just a matter of rhetorical spin. Instead, it cuts to the heart of the legitimacy of the European project and its political autonomy.

Thus, it's no surprise that Commissioner Virkkunen is doing the rounds in the press, fighting off the perception of backtracking under American coercion. If Europe admits to bowing under the pressure of the US, its pride and credibility take a hit. No longer could it claim to see eye to eye with great powers like China and the USA. Europe’s prized regulatory autonomy would be exposed as a mirage. Hence, policymakers feel pressure to save face and feign business as usual—even when in the background, holes in existing rules such as the AI Act are already stretched wide open to create ‘flexibility’ and blunt the law without changing its letter.

Admittedly, the writing has been on the wall for years. In Brussels, the sense of regulatory fatigue is palpable. Business-friendly politicians and tech lobbyists were skeptical of any regulation all along. But now, even those who supported and still support potent rules to govern technology seem disillusioned. It is one thing, after all, to write ambitious laws—and quite another to operationalize the rules and enforce them effectively with limited resources. This problem has surfaced especially in domains that are highly complex and dynamic and in which no one knows what the future holds. AI is a prime example.

The AI Liability Directive fell victim to this waning regulatory appetite. The Commission's decision not to pursue it further is less a sign of coordinated deregulatory vigor and more of a simple inability to gather the necessary troops to push it forward—it is a sign of dissensus, not consensus. Indeed, Europe has traditionally lacked a coherent and forceful tech policy beyond crafting rules. If anything, US dominance in European digital markets has kept on growing. A genuinely coordinated pushback across the tech stack is nowhere to be found; even though the Eurostack concept was hoped to galvanize disparate efforts, competencies remain fragmented across the continent’s governments.

That may be changing. At the AI summit in Paris, US Vice President JD Vance lashed out against “onerous international rules” foisted on American companies amidst tariff threats from President Trump. The new US administration’s belligerence has convinced many in Brussels that EU reliance on US tech is a genuine vulnerability. Crucial digital infrastructure is run by companies like Microsoft. What if Trump forbade further security updates to European Windows PCs unless Brussels follows his lead? If nothing else, digital dependency holds the potential for extortion.

This makes the building of European capacities a matter of crucial importance. So, although looser rules may look like kowtowing to US companies, they may, in fact, be aimed at European firms to entice them into a nascent EU digital industrial policy arrangement. While European industrial giants pledged a 150 billion euro AI investment under the Paris AI Summit, tech CEOs huddled together with European heads of state discussing how “to create a drastically simplified AI regulatory framework.”

At the moment, three forces tug at EU AI policy, and they point in different directions. First, the champions of tech competitiveness still presuppose a world in which global success hangs on beating other companies on price or quality, calling for deregulation and deepening of the markets. In contrast, the crusaders for digital sovereignty take a different angle: they believe only some form of developmentalism can end excessive reliance on US firms—whether through “buy European” clauses in procurement or security-sensitive sectors or other measures.

While these approaches vie for dominance, a third force quietly saps energy from both of them: the inertia and entropy typical of the Brussels machinery. In spite of everyone’s good intentions, forging a united front remains a challenge. When the member states erected EU institutions, they consciously avoided anything like a forceful executive. In times of peace, that served national capitals just fine. Now, that inertia by design comes back to haunt them.

Whether a neo-liberal or hawkish-realist perspective will win out—or no coherent policy at all—remains to be seen. For anything forceful to happen this year or the next, digital sovereignty must remain a Chefsache—a top priority in the EU’s highest political echelons. The White House’s push to impose penalties on foreign governments that tax or fine US tech companies will be the next test: will Europe cave in and defang its policies, or double down on its right to sovereign digital policy?

In the meantime, Trump’s defense of Putin throws Europe into yet another crisis, shifting attention away from the French AI summit, which already feels like years ago. The biggest threat now is an appeased and emboldened Russia, not a technologically dominant USA. Funding for additional howitzers may become more important than additional hyperscalers. To the degree that a digitally sovereign Europe can only emerge with constant coordination and push from the very top, the project might falter yet. At the very least, expect a rocky ride.