Technology announcements come one after another every week: a new AI model, a cybersecurity standard, a change to European regulations. Keeping up with digital news isn’t about collecting headlines; it’s about understanding what’s actually changing in the services we use every day. This article highlights three areas that deserve particular attention as the year draws to a close.
AI Act: transparency requirements already in force
Have you noticed that some chatbots now display a warning such as “this content was generated by AI”? This isn’t a marketing initiative. Since August 2026, the European AI Act’s transparency rules have been in force. They require providers to clearly identify content created or modified by artificial intelligence.
The original timeline called for a gradual rollout. It was revised for so-called “high-risk” systems (AI-assisted medical diagnosis, credit scoring, automated recruitment), but the transparency requirements were not postponed. In practice, this means that any software publisher incorporating generative AI into its product must inform end users now.
E-novateur’s news coverage regularly features explainers on these regulatory developments, which affect both major platforms and small and medium-sized businesses developing digital tools.
The key takeaway: transparency is no longer a voluntary commitment; it’s a legal requirement. Businesses using AI-powered services should check that their providers comply with these new requirements.

Cybersecurity and the NIS2 Directive in France: what’s stalled and what’s moving forward
The European NIS2 Directive was intended to strengthen the digital security of a wide range of organizations, far beyond critical infrastructure operators. In France, its transposition remains on hold pending the parliamentary timetable. The bill is making its way through the process, but delays are mounting.
That doesn’t mean nothing is happening. ANSSI (the French National Agency for the Security of Information Systems) has introduced a pre-registration system and published a Référentiel Cyber France, presented as a working document to help organizations prepare.
Why subcontractors are already affected
Even without a final French law, cybersecurity requirements are already flowing down the contractual chain. A client subject to NIS2 in another EU country may require its French service providers to meet the relevant security standards. Software publishers and hosting providers are among the first to be affected by this mechanism.
In practice, an IT service provider working with German or Dutch clients may already receive contractual clauses aligned with NIS2. Waiting for France to transpose the directive before taking action could mean losing business.
- Check whether your European clients or partners are already applying NIS2 in their countries
- Consult ANSSI’s Référentiel Cyber France to anticipate the expected technical measures
- Identify the sensitive data you process for clients and document how it is protected
AI built into digital services: who pays when things go wrong?
Cloud, CRM and office suite providers are rapidly adding generative AI to their offerings. The feature is often enabled by default or offered as a “free” add-on. This apparent lack of cost obscures a question few users ask: who is liable when an integrated AI tool produces an incorrect or harmful result?
The blurred line between provider and user
Consider a simple example. An accounting software package adds an AI assistant that automatically generates tax summaries. If a summary contains an error that leads to a tax reassessment, is the software provider liable? Or is the user, because they approved the document?
Most software publishers’ terms of service explicitly disclaim liability for AI-generated content. Users remain legally responsible for what they publish or submit, even if it was written by a machine.
The AI Act requires transparency, but it does not yet settle the question of civil liability for general-purpose systems. This gap creates a gray area that businesses must manage themselves by putting human review procedures in place.
The hidden cost of “free” AI
When a provider builds AI into its offering at no visible extra cost, the cost is shifted to the user in the form of risk: the risk of inaccurate data, regulatory non-compliance, or data leaks if prompts are used to train a third-party model.
- Read the terms of use specific to the AI features in your tools
- Set up a systematic human review before distributing any AI-generated content
- Ask your provider where data submitted to the AI is processed and whether it is used to train a model

Digital news and innovation: how to cut through the noise
The volume of tech news published every day makes it difficult to separate signal from noise. French specialist media cover artificial intelligence, cybersecurity and the digital economy, but their approaches vary considerably depending on their audiences.
For non-specialist readers, the most useful way to assess each announcement is to ask three questions. Does it change a rule that applies to my business? Does it affect a tool I already use? Does it create a new risk to my data?
If the answer to all three is no, the information may be interesting, but it isn’t a priority. This approach helps distinguish technological novelty from genuinely transformative innovation. A regulation coming into force matters more than a spectacular prototype.
The coming months will be shaped by the practical implementation of the AI Act, progress on NIS2 in France, and the growing number of AI features in everyday tools. Following these three developments makes it possible to anticipate digital changes and make informed decisions, rather than simply react to them.



