The EU should cut actual red tape, not AI safeguards

Right now, independent experts are working with the EU AI Office on practical guidelines for general-purpose AI companies. As Turing Award-winning computer scientist Yoshua Bengio points out, this effort aims to turn regulatory principles into clear practices that work across all EU countries, preventing a patchwork of different rules in each member state. Under the EU Artificial Intelligence Act, this toolbox will in practice be used by global behemoths like Meta and Google and doesn’t have much to do with European startups. Hence, keys to unlocking actual EU economic growth are not hidden in the AI-safety box at all. It is all about the plain old business bureaucracy that we need to cut.

What the EU should fix

In Estonia, where we are from, launching a company takes just 10 minutes online. A digital form needs to be completed and a fee of roughly €200 paid to the government. For a founder in Belgium, despite identical EU-level rules, bureaucratic hurdles start with over €1,000 in notary fees, mandatory bank account setup that requires in-person meetings, and other requirements that could drag on for weeks. Or if you start in Greece or Italy, for months. These are the things that the EU needs to fix inside member states. 

Furthermore, if you start that business in Estonia, it should be straightforward to expand to any other EU country, across a true digital single market. Instead, we face fragmentation caused by national-level administrative requirements. Hiring, raising capital, and serving customers becomes much more difficult when you cross national borders even inside the EU. More harmonized EU-level standards could help ensure documentation from one member state would be recognized throughout the Union.

As Columbia law professor Anu Bradford notes, the EU lacks a true digital single market, has shallow and fragmented capital markets, punitive bankruptcy laws, a culture of risk-aversion, and an inability to harness global talent. The Mario Draghi report from last year identifies fragmentation, insufficient investment, and regulatory complexity as key barriers, and calls for investment and regulatory harmonization.

This kind of administrative inconsistency, rather than EU digital regulations, adds unnecessary friction to creating and scaling more European tech companies. Yet, in the laudable drive to pull back on regulations to boost entrepreneurship, we’ve noticed a worrisome movement of trying to bundle recommendations that don’t help this objective.

AI giants

There are about 10 companies worldwide building the most capable general-purpose AI models, the ones covered by the safety and security requirements in the EU’s regulatory framework. These are not garage startups. They are some of the most well-resourced corporations in human history. They can handle the requirements set forth in the EU AI Act.

One key requirement is third-party evaluations—meaning that an independent entity tests whether an AI model is safe—under certain conditions. These assessments are designed to ensure that models like Meta’s Llama or Google’s Gemini do not pose systemic risks such as cyberattacks, bio-risk, and loss of control. Claims that such tests would be too burdensome or unworkable is ludicrous given the resources of the companies involved.

Such claims also contradict the prior commitments of top companies at the 2023 UK AI Safety Summit, the 2024 Seoul Summit, and as part of the White House Commitments. 

Getting rid of such AI testing to help EU innovation is like hoping that eliminating checks for hazardous material in Chinese toys automatically helps the EU produce and export more toys for children. Neither addresses the actual barriers to bringing more European technologies to the world.

Sensible AI oversight

Indeed, the safeguards mirror established practices in pharmaceuticals, aviation, and finance where independent verification builds essential public trust and investor confidence. For example, in pharmaceuticals, new drugs must undergo rigorous clinical trials overseen by independent regulatory bodies like the FDA before reaching consumers. Why should AI, possibly one of the most consequential technologies of our time, be exempt from such oversight?

A recent survey by researchers from Stanford University, KU Leuven, the Centre for the Governance of AI, and the Future of Life Institute showed strong expert consensus that third-party assessment of general-purpose AI models is both effective and feasible. Between 62% and 87% of experts agreed it helps reduce systemic risks. Nearly all said it was technically feasible.

Digital regulations are unlikely the main reason behind the EU’s relative lack of technological innovation, and cutting AI safety requirements on big corporations isn’t the solution to it. Robust regulations for powerful digital tools developed by well-funded companies are much needed to increase the trust in this technology in an already risk-averse culture. We can also simultaneously remove unnecessary bureaucratic hurdles for small businesses looking to scale up and fix many of the other culprits impeding business growth along the way.

While the U.S. and China clash over perceived and real AI dominance, the EU has an excellent opportunity to build technology companies rooted in our humanistic values. Let’s not squander it by confusing safeguards with red tape and losing sight of the real obstacles to innovation.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Read more:

• Trump has inadvertently shown Europe it needs to build a full-stack AI industry—and avoid a risky reliance
• Trump is killing the goose that laid America’s golden eggs
• Europe’s tech sector will benefit from Trump disrupting markets—if it seizes the opportunity, says early Spotify investor