In response to the rapid development of artificial intelligence, governments are enacting laws to regulate it, such as the European AI Act.
There are high expectations for this legislation, which will gradually set guidelines for the ethical use of AI within the European Union.
Artificial intelligence is no longer a distant promise but an operational reality that is profoundly transforming the economic and social fabric. However, this technological acceleration is accompanied by a growing need for ethical and regulatory guidelines. It is in this context that Europe has drafted the AI Act, an ambitious regulation aimed at governing the use of AI.
To understand its practical implications, we spoke with Mathieu Changeat, co-founder of Dydu and now chief operating officer of the Zaion Dydu Group. His analysis, grounded in practical expertise and a sociological perspective, reveals a complex landscape where sovereignty, trust, and business pragmatism vie for center stage.
The AI Act: An Essential Legislative Framework, Not a Panacea
The AI Act, discussions of which predated the ChatGPT boom, is a European response to the need to clarify the use of AI. It will therefore impose obligations regarding model training and protect intellectual property.
Its implementation was initially scheduled for 2026. It will now be phased in gradually through 2028, but this delay should not be interpreted as a sign of weakness. According to Mathieu Changeat, Europe’s real challenge lies more in a massive lack of funding compared to the American and Chinese giants.
“The AI Act was supposed to be the big bang of August 2026, but there’s already talk of delays until 2028 for certain aspects. But I think the problem lies elsewhere. It’s funding,” he explains.
Indeed, despite the emergence of European players valued at billions—notably Mistral AI—the financial firepower is still lacking. It will be necessary for the development and democratization of generative AI models.
However, the AI Act plays a crucial role as a certification label. The “AI Act” compliance label will strengthen users’ perception of trust—even though it does not guarantee the total absence of algorithmic hallucinations. These are a result of the algorithms’ technical architecture. The label certifies that the models have been trained on ethical data, thereby enhancing reassurance and transparency.
Sovereignty and the Digital Divide: The European Challenge
One of the most significant impacts of the AI Act is evident in the quest for digital sovereignty. Mathieu Changeat observes a shift in mindset, particularly among large corporations and the public sector. He notes that “for the public sector and large enterprises, digital sovereignty has become a priority, sometimes at the expense of pure performance. They now want European solutions in light of the current geopolitical situation.” ”
Digital sovereignty is becoming a major reassurance factor for large enterprises and the European public sector. As a result, they are favoring local solutions even in the face of initial performance gaps. This trend toward self-regulation, a precursor to the AI Act, means that the Act will not disrupt their established practices.
However, a technological divide is emerging. While large corporations have the resources to invest in customized, sovereign solutions, questions remain for SMEs. They risk becoming more dependent on off-the-shelf solutions, often from the U.S. The reason: a lack of resources and time to keep pace with regulatory changes.
Risk Management and Ethics in the AI Act
The advent of autonomous AI agents, capable of collaborating on complex tasks, raises new governance issues. Mathieu Changeat emphasizes the need to strictly define the agents’ scope of action and to implement technical safeguards.
The goal is to prevent unexpected behavior or problematic interactions. Similarly, this approach remains more accessible to large corporations capable of financing these custom developments.
Furthermore, user mistrust regarding their personal data is rising sharply. Studies indicate that the cookie refusal rate will reach nearly 40% by 2026. This heightened vigilance is pushing companies to anticipate transparency requirements, even before the AI Act is fully implemented.
The main “red flag” for a company remains the “black box” nature of certain AI models. As projects evolve, systems that were initially stable may become incompatible, leading to contradictory results that could be penalized under the AI Act. For customer service, the use of AI is not “high-risk” as long as transparency is maintained and access to data is limited to what is strictly necessary.
Humans Augmented, Not Replaced: The Customer Service Paradox
Forecasts warn daily of the risk of humans being massively replaced by machines as AI becomes more widespread. However,contrary to alarmist predictions of massive automation in customer service, Mathieu Changeat advocates for a more nuanced view.
“90%: that’s alarmist; I don’t really believe it based on my experience. Ultimately, it’s people’s choice: if we give users the option to be connected with a human agent or a chatbot, the majority choose the human. ”
Indeed, users still prefer human interaction for complex situations. These are the situations that require empathy and a nuanced understanding of the context. AI is more likely to enhance the capabilities of human agents.
It will therefore relieve them of tedious tasks, allowing them to focus on high-value-added interactions. However, it’s important to note that since the AI Act is not a labor code, it will have no direct impact on protecting employees from automation.
Outlook for 2028: The AI Act as a Reassurance Label, Not a Constraint
Looking ahead to 2028, Mathieu Changeat estimates that AI will be standardized. The AI Act will not represent a restrictive “big bang” for companies already familiar with best practices in data management (GDPR). Rather, it will be part of a continuum of regulation, serving as a seal of reassurance for users.
Mathieu’s ultimate strategic advice is clear. We must focus on real business needs rather than succumbing to the AI fad. Many use cases do not require complex tools to be solved, and a pragmatic, value-added approach remains the key to success.
The AI Act, therefore, does not present itself as an insurmountable barrier to innovation. Rather, it will act as a catalystfor more ethical, transparent, and sovereign AI. While funding challenges and the technological divide between large corporations and SMEs remain significant, awareness and self-regulation are already underway.
Will the future of AI in Europe thus take shape as a model of technological development that serves people and fosters trust, balancing performance with social responsibility? Only time—and the ability of stakeholders to incorporate these imperatives—will tell.
Article published by lebigdata.fr
Share this article!
