The Governance Illusion: How AI Illiteracy Is Costing Europe – and Leaving the Real Risks Unaddressed (Part 3)

When Regulation Becomes a Competitive Disadvantage

Part 3 of the series: The Governance Illusion

So far, we’ve looked at the technical illiteracy baked into the EU AI Act — a flawed AI definition, untrackable agentic pipelines, transparency that doesn’t actually inform, and incident reporting that can’t detect patterns. But there’s a second tier of damage that deserves its own honest reckoning: the economic cost of getting governance wrong.

Because when regulation is poorly designed, it doesn’t neutrally fail. It actively shifts the playing field — away from the companies and regions that tried hardest to follow the rules.

Two Diagnoses Europe Can’t Ignore

In 2024, two of Europe’s most respected economic statesmen issued independent diagnoses of the continent’s competitiveness. Neither pulled punches.

Mario Draghi’s September 2024 report on European competitiveness arrived with the kind of authority that makes policymakers uncomfortable. His findings on regulation were stark: the EU now has around 100 tech-focused laws and over 270 regulators active in digital networks across Member States. More than 60% of EU companies see regulation as an obstacle to investment. For SMEs, the number rises to 55% — and for them, it’s not an abstract concern, it’s an existential one. On AI specifically, Draghi’s report noted that the AI Act’s computational thresholds for systemic risk — the very thresholds we examined in Part 2 — were already being exceeded by state-of-the-art models at the time of writing. His prescription? An AI Sandbox regime that would regularly assess whether EU and national legislation is getting in the way of innovation. That is not the language of a man confident that the current framework is working. 

Enrico Letta’s April 2024 report on the EU Single Market took a gentler tone — welcoming the AI Act as a “decisive step” toward a modern digital strategy, while simultaneously calling for a “fifth freedom”: free movement of knowledge, research, and innovation. Read both paragraphs together and you spot the tension immediately. Praising the Act in one breath, identifying the very scalability failure the Act contributes to in the next. The authors simply lacked the technical framework to connect those two observations. 

That gap — between what regulators intend and what they actually understand — is the through-line of this entire series.

What Does Compliance Actually Cost?

Numbers matter here, so let’s be precise. Several independent studies have attempted to quantify the EU AI Act compliance burden.

Now ask yourself: who can absorb those costs? The Draghi report answers clearly — only larger companies, often non-EU based, have the financial capacity to comply at scale. Young innovative European companies may simply choose not to operate in the EU at all.

The Market Has Already Voted

If you want to know whether a regulatory framework is working, watch what companies actually do — not what they say in consultation responses.

Meta delayed its EU AI rollout in June 2024 after Ireland’s Data Protection Commission intervened on GDPR grounds, blocking LLM training on EU user data. When Meta finally launched in March 2025, it was a degraded product — trained without EU data, with weaker regional language and cultural adaptation. Meta’s description of the EU regulatory landscape? “Complexity, inconsistency, and uncertainty.” 

Apple delayed the EU launch of Apple Intelligence, iPhone Mirroring, and SharePlay Screen Sharing, citing DMA interoperability requirements. The delay stretched into 2025. Apple also declined to sign the EU’s voluntary AI Code of Practice — joining Meta on the sidelines. 

Mistral AI sits in the most nuanced position. Cédric O, former French Minister for Digital Affairs and Mistral co-founder, warned that the AI Act could “kill” Mistral. CEO Arthur Mensch’s public stance is more measured: “On the AI regulation front, we’re in a workable state. It’s not ideal: I think the EU came a little too early on regulation.”  Importantly, Mensch identifies market fragmentation and low enterprise AI ambition — not the AI Act alone — as Europe’s primary AI challenges. That’s a materially important distinction, and intellectual honesty demands we note it. 

The Investment Gap Is Real

Here’s the number that should keep European policymakers up at night: the EU represents roughly 17% of global GDP but attracts only 7.5% of global AI investment. 

Regulatory burden is not the only explanation for that gap. Europe also has structural issues — fewer large institutional investors willing to take early-stage AI bets, a fragmented single market that makes scaling harder, and cultural risk aversion in enterprise AI adoption. Mensch is right to name those.

But regulatory burden is an avoidable factor. And avoidable factors that cost €3.3 billion a year in compliance overhead, drive product withdrawals, and deter young companies from entering the EU market deserve direct, uncomfortable attention — not diplomatic hedging.

The Honest Question

Europe wants to be a global leader in trustworthy AI. That is a genuinely worthy ambition. Responsible AI governance is not optional — it is necessary for protecting citizens and building long-term trust in the technology.

But trustworthy AI governance requires two things to be true simultaneously: it must be rigorous enough to address real risks, and technically literate enough to target those risks accurately. Right now, the EU AI Act achieves neither. It imposes heavy costs on the wrong actors, leaves real risk vectors unaddressed, and is already being outpaced by the technology it was designed to govern.