Commentary. Europe’s artificial-intelligence debate has entered its most consequential phase: the rules are no longer merely being drafted, but implemented, revised and tested against political pressure. Our view is straightforward. The European Union should simplify the AI Act where compliance is needlessly confusing, but it should not mistake delay for pragmatism or weaker safeguards for competitiveness.
The AI Act began as an unusually ambitious attempt to regulate technology according to risk. Its most demanding obligations target systems used in areas such as recruitment, worker management and access to education, where an opaque automated decision can materially alter a person’s life. The framework is being enforced in stages, while the EU’s broader platform-work rules are due for national implementation by December 2026.[3][9]
That timetable has created legitimate concerns. Companies—especially smaller firms—argue that overlapping obligations, uncertain technical standards and the cost of documentation could discourage investment or push innovation elsewhere. European policymakers have responded with a digital “omnibus” approach intended to simplify parts of the framework, and some high-risk obligations have been moved to later dates.[3][9]
Those concerns deserve more than dismissal. A law that only large technology companies can navigate may entrench the very incumbents it claims to constrain. Regulators should eliminate duplicated paperwork, publish usable compliance templates and give smaller developers access to testing facilities and legal guidance. Enforcement should be predictable, proportionate and focused on genuine harm rather than formal mistakes.
But simplification becomes dangerous when it quietly changes the substance of the law. The people most exposed to automated decisions are rarely in a strong position to challenge them. A rejected job applicant may never know an algorithm shaped the decision. A worker managed by software may find that an appeal process exists on paper but not in practice. Transparency, human oversight and the ability to contest a consequential decision are therefore not luxuries reserved for a slower economy.
“The rules are no longer merely being drafted, but implemented, revised and tested against political pressure.”
The United States illustrates the other risk: fragmented protection. In 2026, American states continued to legislate on chatbots, automated decisions, frontier-model risks and third-party auditing, while no federal AI bill had passed during the current session.[5] State experimentation can produce useful policy, but a patchwork also raises compliance costs and leaves protections dependent on geography.
Europe should not respond to American fragmentation by abandoning its own standards. Nor should it assume regulation alone will make AI trustworthy. Rules must be backed by technically capable authorities, meaningful penalties and enough staff to investigate companies that conceal failures. The Commission says enforcement by the AI Office and national authorities began on 2 August 2026, but the transition remains administratively demanding.[9][10]
The strongest case against overregulation is not that accountability is unnecessary; it is that badly designed accountability can make beneficial tools harder to deploy. AI could improve accessibility, scientific research and public services. Excessively cautious rules may protect institutions from reputational risk while denying citizens useful innovation. That is why evidence-based review matters, and why regulators should distinguish experimental, low-risk applications from systems that decide who gets work, credit, education or freedom from surveillance.
Yet the strongest case for firm enforcement is equally practical. Trust is infrastructure. Businesses and citizens will adopt AI more readily when they know its limits are real, its outputs can be challenged and its makers cannot simply shift responsibility to an automated system. A permissive market may deliver rapid deployment, but a market without credible red lines invites scandals that can set innovation back further than sensible regulation ever would.
Our position is not “regulate everything.” It is to regulate consequences. Europe should streamline the AI Act’s procedures, maintain its protections for high-impact uses and publish clear evidence whenever obligations are postponed or weakened. Democratic oversight should not be treated as an obstacle to competitiveness. It is the mechanism that can make technological progress durable.