The European Union is recognised as a global pioneer in the regulation of artificial intelligence, thanks to one of the most comprehensive legislative frameworks – the EU Artificial Intelligence Act (AI Act).
The text entered into force on 1 August 2024, but the key requirements for high-risk systems and general-purpose "GPAI" systems will only begin to be implemented in the coming years.
With this law, the EU aimed to strike a balance between innovation and the protection of fundamental rights.
It was intended to be a European moment of determination – for Europe, while the rest of the world pursues innovation without limits, to show that human rights and privacy are not collateral to progress.
From ambition to caution
But today, just one year later, the European Commission is considering how to "soften" that same law. A leaked document shows that Brussels is preparing the so-called Digital Omnibus—a package of changes that, according to the draft, is due to be published on 19 November. It contains proposals to delay, mitigate or redefine certain obligations arising from the AI Act.
At the root of this is the fear that Europe has gone too far. The Commission is increasingly convinced that implementing the law too quickly could lead to an outflow of investment and a decline in the competitiveness of the European technology industry.
Meta, Google, and Microsoft, say that Europe faces the risk of "stifling its own innovation impulse"
At the same time, pressure from the United States has become direct. Donald Trump's administration claims that European regulations "unfairly target American companies", and threats of possible trade measures are being made in the background.
Silicon Valley tech giants, such as Meta, Google, and Microsoft, say that Europe faces the risk of "stifling its own innovation impulse." Their message is simple: give us time and space.
Inside the Digital Omnibus draft
The Digital Omnibus draft suggests delaying penalties for generative AI systems, giving a one-year "grace period" to developers and sellers of models like ChatGPT or Claude.
This implies that provisions scheduled to be enforced in 2025 could be delayed until at least the second half of 2027.
The Commission is also thinking about allowing some high-risk systems, especially those used for administrative and procedural tasks, to avoid the stricter rules for registration and supervision.
More seriously, changes to the application of rules on the processing of personal data for AI model training are under discussion.
The new draft would permit companies to invoke "legitimate interest" even when processing sensitive categories of data, which was previously expressly prohibited.
"A death by a thousand cuts" - Max Schrems (NOYB)
Privacy and digital rights advocates argue that this sets a dangerous precedent. The Austrian organisation NOYB (None of Your Business), led by prominent activist Max Schrems, called it "a death by a thousand cuts"—alluding to the fact that small changes in practice result in a systemic weakening of the GDPR.
The European Parliament, which must approve any changes, already has MEPs who are openly speaking about "betrayal of the European principle of digital sovereignty".
The EU retreats before the market
Legally, the AI Act is designed as a mechanism to classify systems by level of risk. At the lowest level are those with minimal risk – ordinary AI tools, chatbots, and recommendation algorithms.
At the top of the pyramid are systems that directly affect people's lives: employment algorithms, banking decisions, medical diagnostics, surveillance, and justice.
These systems are subject to strict supervision and must meet clear standards of transparency, human oversight, and security testing. The principle was that responsibility could not be delegated to a machine. It is precisely in this most sensitive area that "flexibility" is now being sought.
The European Union now appears to be retreating in the face of market interests
Officially, the Commission claims it is not abandoning the goals of the AI Act but is seeking a way to avoid "overburdening" the economy. In informal conversations with journalists, European officials refer to the "rationalisation" of regulations and the "alignment" of the AI Act with GDPR and the ePrivacy Directive.
However, behind the bureaucratic language lies a fundamental shift in political tone. The European Union, which has built its reputation as a regulatory force over many years, now appears to be retreating in the face of market interests.
Losing moral authority
If the proposals are adopted, companies will have additional time to adjust their models. AI investment in Europe is likely to increase as legal uncertainty diminishes. At the same time, however, Europe risks losing what it was known for: its moral authority in protecting privacy and human rights in the digital sphere.
Confirmation of the Digital Omnibus in its current form could give Europe more breathing room in the technology race
For users, this could mean less control over how their data is used to train algorithms. For small and medium-sized European companies, which have so far relied on clear rules and a transparent framework, it may mean increased uncertainty.
If the "rules of the game" change midway through implementation, those without lobbying power suffer most.
From a global policy perspective, this is much more than a technical issue. The European Union was the only power to shape the digital framework through laws with global reach. When the EU gives in, it does not remain merely an internal matter.
China and the US, which are building their AI development models on entirely different principles, will gain a new argument: if Europe is easing, why should anyone remain committed to regulation?
Europe at a turning point
The digital world is at a point where the speed of innovation is measured in months, while legislation is measured in years. That gap is increasingly difficult to close. What seems pragmatic today may become an abandonment of principle tomorrow.
That is why 19 November, 2025, will be an important day—not only for the European Commission but also for the idea that Europe is the "regulatory compass of the world."
Confirmation of the Digital Omnibus in its current form could give Europe more breathing room in the technology race.
But it will lose some of the trust it has built up over the years – the trust that the power of technology can be contained by law, not by compromise.
