Those high-level conversations about deploying artificial intelligence responsibly have so far gone ahead without a seat reserved for Europe — a gap that stands out given that its legislators were the first in the world to enact a broad regulatory law governing the technology. Europe's position in artificial intelligence development trails that of both the United States and China, according to what this episode brings into view.

Having moved earlier than most regions to regulate artificial intelligence, Europe nonetheless assembled a comparatively narrow foundation of frontier models, data infrastructure, and large technology companies — a gap that economist Philippe Aghion, a Nobel laureate, assessed as leaving the continent at real risk of being pushed to the margins as AI transforms the architecture of global competition. He co-authored a report calling on Europe to adjust its direction on artificial intelligence if the region hopes to remain competitive in the technology's development.

Around the same period that a predominantly American cohort of AI industry leaders pledged voluntary oversight of the field's top-tier models, the regulatory path taking shape reflected little of what the EU had put forward. Building technology companies at the scale seen in the United States has proven an uneven challenge across European markets, where no homegrown firm has yet matched the reach or valuation of the largest American platforms. No domestic AI tools have yet matched the capabilities demonstrated by models such as OpenAI's ChatGPT, Anthropic's Claude, or DeepSeek's offerings from China.

In a speech delivered the previous month, ECB president Christine Lagarde characterised the position Europe finds itself in on artificial intelligence as uncomfortable, outlining a tension between two paths: restraining adoption to safeguard data sovereignty at the cost of economic growth, or embracing AI rapidly and accepting deep dependency along with the attendant risk of ceding control over how the economy is shaped by its own principles.

Mistral AI, the Paris-based startup, released a new model last week that the company positions as comparable to those of its American counterparts — a development that offers some measure of optimism for the European AI sector, though whether it can mount a sustained challenge to the established players in the field remains an open question.

Long before the governance of artificial intelligence tools became a widely contested policy question, the EU's AI Act took shape as a framework designed to make companies answerable for the risks their products generate — covering obligations around data quality, human oversight, and cybersecurity. Penalties for falling short of the requirements range from mandated adjustments to tools all the way up to financial levies reaching 3% of a company's yearly revenue — a threshold that translates to billions of dollars when applied to the industry's largest players.

Described by some as a key instrument for establishing oversight boundaries around artificial intelligence, the AI Act has drawn attention as regulators look to define the boundaries of responsible development. Von der Leyen, delivering her state of the union address the previous month, argued that the bloc was now positioned to help steer worldwide conversations about managing the risks posed by the technology.

Putting the law into practice has run into obstacles, with a number of EU member states yet to put oversight bodies in place to monitor adherence. Together with fellow tech frameworks such as the Digital Markets Act and the General Data Protection Regulation, this legislation has played a part in slowing the introduction of artificial intelligence products across European markets.

Meta AI's European debut came in March 2025, roughly a year and a half after the platform first launched in the United States, with the company noting at the time that working through the region's regulatory framework had pushed the rollout beyond its original timeline. In a Eurostat survey of around 157,000 businesses conducted the previous year, approximately half pointed to uncertainty over legal exposure and the prospect of running afoul of the EU's data protection framework as factors behind their decisions to hold off on deploying AI.

Eléonore Crespo, who co-founded Pigment — a Paris-based business-software company carrying a valuation above $1 billion — told a broadcaster that the AI Act and related national frameworks were hampering her firm's pace, and that such constraints would ultimately undermine the sector's ability to function. Regulatory compliance costs have been a point the European Commission itself put on record.

Passed through the legislature the previous year, the AI Omnibus bundled together a set of provisions that softened certain obligations under the AI Act and pushed the deadline for rules governing high-risk AI systems out to December 2027.

Observers spanning legal scholarship and AI research have noted that any verdict on the AI Act's effectiveness remains premature, given that its provisions are still being rolled out in stages — a view echoed by Amir Banifatemi, who holds the role of chief responsible AI officer at Cognizant AI lab, a San Francisco-based research organisation, and who told a broadcast outlet that while the act's prohibitions rest on broadly reasonable foundations and its transparency obligations are relatively manageable, the framework governing high-risk applications represents its most vulnerable area and has yet to come into force. Banifatemi drew attention to AI agents as an area worth examining, observing that their tendency to draw on tools built by multiple providers leaves the matter of liability — should anything go wrong — without a clear answer.

Data compiled by Stanford University's Institute for Human-Centered Artificial Intelligence put European private-sector AI funding at around $21 billion over the course of that year, set against a United States figure of $285.9 billion drawn from the same period. According to figures published in the AI Index Report from that centre, data centre infrastructure in the United States outnumbers any single rival nation by a factor of ten, while 59 AI models of note originated there across 2024 — a tally that placed it well ahead of China's 35, with France and the United Kingdom each contributing one.

Crespo described the continent's current position in artificial intelligence as resembling an early, underdeveloped era, while expressing confidence that a period of broader advancement lies ahead, pointing to promising projects as grounds for that view. Banifatemi, meanwhile, argued that deep industrial knowledge and sector-specific data across manufacturing, energy, transport, and healthcare place the region in a strong position to capture much of the economic value AI may ultimately generate.