The European Commission has decided to give a major boost to big technology companies. This is what emerges from the digital regulation simplification project presented on Wednesday in Brussels, whose declared objective is to ensure that companies dedicate less time to bureaucratic procedures and more time to innovation. One of the main innovations included in the project, which both member states in the Council and the European Parliament have yet to validate, is to delay the application of the ban on high-risk artificial intelligence (AI) systems “for up to 16 months”. The grace period is even longer than the year considered, as reported by EL PAÍS, in the preliminary versions of the project.
The European AI regulation, in force since August 2024, although its full implementation was expected two years later, classifies AI systems according to the risks that their application poses for citizens and, based on this, assigns them various obligations and requirements to comply with. Therefore, harmless risk tools, such as spamthey have no restrictions, while those with unacceptable risks are directly banned. That category includes those “that transcend a person’s consciousness or deliberately manipulative techniques,” those that exploit a person’s vulnerability, or those that infer people’s emotions, race, or political opinions.
Below the prohibited ones are the high-risk applications, which require permanent supervision. These include remote biometric identification systems, biometric categorization systems, those affecting the security of critical infrastructure, and those related to education, employment and the provision of essential public services, law enforcement or migration management.
The oversight that these high-risk AI systems should be subjected to is one that is now proposed to be delayed by up to 16 months: instead of starting to apply in August 2026, as initially planned, the deadline is extended until December 2027. Brussels’ argument to justify this concession is that, before enforcing compliance with the standard, standards must have been published on what is and what is not acceptable for these tools to do.
“This is not a delay in the regulation of AI, which is already in force,” underline community sources, who also reject that the so-called digital omnibus law is a concession to technology companies. The objective is and remains to “fully apply it, respect the law on AI”, they insist. But this requires there to be complete “legal certainty” about how the rules are applied, and that is what the re-establishment of the EU’s digital rules aims to provide, they argue.
“If the plan is implemented, the Commission will end up bowing to the pressure of Great technology in almost everything they asked for”, warns former MP Ibán García del Blanco, in reference to the constant pressure exerted in recent months by the large American – but also European – digital platforms and, above all, directly by Donald Trump’s government, which went so far as to define European legislation on AI as “excessive”. “It seems more like a lowering of security standards than a simplification of processes. AI regulation already involved a delicate balance between technology and rights protection. “This is not good news for the European model of human-centered AI governance,” adds García del Blanco, who was one of the negotiators in the European Parliament of the law that is now being amended.
Data and cookie policy
The project to reformulate the digital legislation presented today also includes innovations on the data protection front. The Commission’s intention is to “promote additional measures for (the training of) artificial intelligence models”. These measures will increase access to this data, for example through “data laboratories”, a figure that remains to be specified. Granting additional data to train AI models is justified to try to advance European AI companies.
A modernization of the relevant regulations cookies which aims to improve the browsing experience of Internet users by limiting the number of times the banners From cookies and making it possible to give consent (or deny it) with a single click, and those preferences can be saved.
