EU AI Act’s Testing Gap Leaves People Exposed to Risky AI

Europe’s landmark AI law, the EU AI Act, was supposed to protect everyday people from the dangers of unchecked artificial intelligence. But a new analysis warns that regulators can’t enforce rules on AI systems they don’t know how to test—leaving workers, families, and students at risk from unproven or unsafe technology.

What Happened

On October 4, 2026, Cybersecurity Insiders reported a significant shortcoming in the European Union’s AI Act: regulators lack reliable ways to test or fully understand many advanced AI systems. The AI Act, which was passed by the European Parliament in 2024 and is set to take effect in 2026, is the world’s first comprehensive law aimed at governing artificial intelligence. It sets out strict requirements for high-risk AI—such as those used in hiring, education, healthcare, and law enforcement—and bans certain uses outright.

But the law’s effectiveness depends on regulators’ ability to verify that AI systems are safe and fair. The problem? Many AI models, especially the latest “black box” systems from companies like OpenAI, Google, and Anthropic, are so complex that even their creators can’t always explain how they work or predict their behavior. Without clear methods to test these systems, the EU’s rules risk becoming little more than promises on paper.

This gap isn’t just a technical issue—it’s a real-world problem that could affect millions. If regulators can’t see inside the AI “black box,” how can they stop discrimination in hiring, bias in school grading, or unsafe decisions in healthcare?

How This Affects Everyday People

For workers, parents, students, and anyone who interacts with AI-powered systems, the stakes are high. The EU AI Act was meant to stop companies from using AI to make unfair or unsafe decisions about jobs, education, or access to services. But if authorities can’t reliably test these systems, some risky AI applications could slip through the cracks.

Imagine a job applicant in France rejected by an AI-powered hiring tool. The law says the system must be fair and explainable, but if the regulator can’t actually test the tool, how can that worker challenge a bad decision? Or consider a student in Germany whose test is graded by an AI system. If the grading algorithm is biased or buggy, but too complex to audit, the student might never know—and have no way to appeal.

Families could face similar risks. AI is increasingly used in healthcare to recommend treatments or flag risks. If a hospital in Spain uses an untested AI system that misdiagnoses a child, parents may have little recourse if the system’s logic is a mystery even to regulators. The same goes for AI-powered surveillance or predictive policing, which can affect entire communities if left unchecked.

In short, the promise of the EU AI Act—to protect people from AI harms—depends on the ability to actually test and understand these systems. Without that, everyday people are left trusting that companies will police themselves, a gamble that history shows rarely works out for the public.

The Bigger Picture

This isn’t just a European problem. Around the world, governments are racing to regulate AI, but many face the same challenge: the technology is evolving faster than the tools to test or understand it. In the U.S., the Biden administration’s 2023 AI Executive Order called for new standards and testing methods, but progress has been slow. Meanwhile, companies like OpenAI and Google are rolling out ever more powerful models—often with little transparency about how they work.

Recent incidents show what’s at stake. In 2024, a major AI-powered hiring platform used by several European retailers was found to systematically reject older applicants, but only after months of investigation. In the education sector, a 2025 study revealed that automated essay grading tools used in UK schools favored certain writing styles, disadvantaging students from diverse backgrounds. These are exactly the kinds of harms the EU AI Act was meant to prevent—but only if regulators can spot them in time.

The EU is not alone in facing public skepticism. Across Europe, protests and backlash against AI have grown, with workers and parents demanding more say over how AI is used in their lives. As covered in our AI backlash explainer, many feel that laws and promises aren’t enough if there’s no way to enforce them.

For a deeper dive into what the EU AI Act was supposed to achieve—and where it falls short—see our guide on what the EU AI Act means for you.

What You Can Do

• Ask questions about AI decisions: If you’re affected by an AI-powered system at work, school, or in healthcare, ask how the decision was made and whether a human can review it.
• Know your rights: Under the EU AI Act, you have the right to an explanation for high-risk AI decisions. Don’t be afraid to demand transparency from employers, schools, or service providers.
• Support organizations fighting for AI accountability: Groups across Europe are pushing for stronger oversight and better testing of AI systems. Consider joining or supporting their efforts—see our section on how people are fighting back against AI.
• Stay informed: Follow news about AI regulation and incidents in your country. The more you know, the better you can protect yourself and your family.
• Share your experiences: If you’ve been harmed or unfairly treated by an AI system, speak up. Public pressure has already led to changes in some cases.

The Bottom Line

The EU AI Act is a landmark law, but its promise to protect people from AI harms is only as strong as the tools regulators have to enforce it. Until governments can reliably test and understand advanced AI systems, some risks will remain unchecked. For now, everyday people need to stay vigilant, demand transparency, and push for stronger oversight—because when it comes to AI, trust without testing isn’t enough.