The European Commission has recently decided to delay enforcement of the high-risk provisions in the EU AI Act. This has stirred understandable frustration across the AI industry, especially among those of us who have invested years in building technology that is transparent, safe, and fair. 

On one hand, it gives naïve and poorly designed AI systems more time to operate, putting fairness and transparency in hiring at risk. On the other hand, the pause may give regulators the breathing room needed to finalise and build the certification framework vendors have been waiting for, potentially strengthening the Act’s long-term impact if that time is used wisely. 

At Hubert, we see this delay as unfortunate. It underscores a deeper truth: the best organisations aren’t waiting for regulation to do what’s right for people.

As a provider of technology designed to make job application processes more efficient and fair, we have long welcomed the EU AI Act. The EU’s ambition to curb irresponsible or naïve uses of AI should be applauded. Just like what GDPR did for personal data, the AI Act has the potential to elevate industry standards and eliminate harmful practices.

Because let’s be clear: questionable uses of AI in recruitment do exist. These include:

• Opaque, unvalidated assessment models making decisions that affect candidates’ lives
• Unreliable or biased scoring systems sold as “AI magic” with no measurable fairness benchmarks
• Non-explainable tools that even vendors themselves cannot interpret
• Automated assessments that prioritise efficiency at the expense of equality and dignity

These are precisely the types of technologies that the EU AI Act seeks to remove from the market, and why Hubert AI has supported the EU AI Act from the beginning.

Delaying the legislation on high-risk practices isn’t ideal. It means that the bad solutions and naïve or sloppy use of AI by organisations, practices which would never comply with the AI Act in the first place, will get two more years to exist. And candidates will pay the price as unfair AI-driven recruitment processes will continue to be around.

The delay also prolongs uncertainty. Will there be further changes? Additional postponements? This ambiguity risks encouraging vendors and hiring organisations to stall rather than prepare.

While businesses may want extra time to adapt, the AI Act has been on the horizon for years, and responsible organisations have already begun preparing. Waiting until enforcement becomes mandatory is simply shortsighted.

But companies are not the only ones to blame; governments have also been remarkably slow in appointing the national competent authorities responsible for overseeing the EU AI Act. Member states were expected to designate these authorities by August 2025, yet many have only done so recently. 

Meanwhile, vendors still lack clear guidance on what constitutes an appropriate or acceptable level of accuracy and robustness for high-risk AI systems. For those of us actively working to comply with the AI Act, operating without clarity on these basic requirements is understandably frustrating.

Still, the next two years offer a chance to correct the course. National bodies and certification organisations now have the time to accelerate the development of clear, actionable standards.

Despite the delays, organisations should not wait to adopt the processes required by the AI Act or to choose suppliers who are already EU AI Act–ready. 

Regardless of enforcement, the proposed regulations should be something to follow because they are common sense for humanity. Explainability and consistency aren’t just regulatory boxes, they are the foundation of responsible AI.

Hubert’s mission remains unchanged.

We will continue to build technology that improves hiring while protecting candidates. We will keep investing in transparency, fairness, and measurable quality. And we encourage organisations to:

• Continue to work with and adopt AI Act–aligned processes now
• Choose vendors that are already EU AI Act-ready
• Use this time to prepare, validate, and future-proof their AI practices
  
  

Because doing the right thing for people should never be delayed.