When a team of developers must choose between migrating an AI model in-house or relying on a third-party API, the answer in 2024 is no longer solely dependent on the budget. It also depends on the European regulation 2024/1689, which came into effect on August 1, 2024, requiring documentation and evaluation of each AI system according to its risk level. This year’s IT trends are shaped by such concrete constraints, not by futuristic promises.
European AI Regulation: What It Changes for Development Teams
There is much talk about generative artificial intelligence, but rarely about the legal obligations that now govern it. Regulation (EU) 2024/1689, published in the Official Journal of the European Union on July 12, 2024, structures the use of AI by risk categories. Applications deemed high-risk (automated recruitment, credit scoring, biometric surveillance) require comprehensive technical documentation, a compliance assessment, and post-deployment monitoring.
For developers and companies, every AI project must integrate compliance from the design phase. It is no longer possible to deliver a model into production without having traced the training data, tested for biases, and formalized governance. Feedback varies on this point: some organizations have anticipated these requirements, while others are discovering them when moving to production.
To keep up with the evolution of these topics, IT on Journal du Webmaster regularly covers the practical implications of these regulations on digital professions.
NIS2 Directive and Cybersecurity: Responsibility Rises to Management
The NIS2 Directive (Directive (EU) 2022/2555) was to be transposed by member states before October 17, 2024. Its scope far exceeds that of the first version. It now covers sectors such as waste management, medical device manufacturing, and postal services, in addition to traditional operators (energy, transport, health).
On the ground, the most concrete change affects the responsibility of executives regarding cybersecurity. Senior management can no longer delegate without oversight: they must validate the risk management strategy, ensure that incident notifications meet regulatory deadlines, and supervise the security of the subcontracting chain.
For IT teams, this translates into several simultaneous projects:
- Mapping all providers and subcontractors with access to information systems, then assessing their security level with formalized criteria.
- Establishing an incident notification process with specific timelines, including a rapid initial report and a detailed report in the following days.
- Documenting the risk management measures adopted and submitting them for validation by management, not just by the CISO.
The term “cybersecurity” appears in all articles about IT trends. The difference in 2024 is that NIS2 transforms a technical subject into a corporate governance obligation.
Data Center Energy Consumption: The Physical Limit of AI
The enthusiasm for generative AI masks a problem that infrastructure teams know well: the energy bill is skyrocketing. In France, data centers consuming more than 1 GWh per year used 4.3 TWh of electricity in 2024. The trend is upward, driven by the computing and cooling needs associated with artificial intelligence models.
Specifically, when deploying a language model or a massive data processing pipeline, the electricity consumption of the data center becomes a top budget item. Technical choices (model size, retraining frequency, geographical location of computing) have a direct impact on costs and carbon footprint.
This constraint is pushing some companies to rethink their approach. Rather than training ever-larger models, we see the emergence of strategies for more compact models, specialized in a specific domain, which consume fewer resources for comparable results within their scope. The choice between a generalist model hosted by a hyperscaler and a lightweight model deployed locally becomes as much a technical decision as an environmental one.
Python, JavaScript, and AI-Assisted Development Tools
On the programming languages side, Python remains the dominant choice for everything related to machine learning, data processing, and rapid prototyping. JavaScript retains its place in web development and frontend applications. Nothing seems new at first glance, but the way of coding with these languages has changed.
AI-assisted development tools (code completion, test generation, vulnerability detection) have integrated into developers’ daily workflows. We are no longer talking about a gadget: the AI assistant changes the time allocation between writing and reviewing code. Developers spend more time validating and correcting suggestions than writing code from a blank page.
This change has practical consequences for the skills sought:
- Knowing how to formulate precise prompts to obtain usable code, rather than a draft that needs to be completely rewritten.
- Mastering critical review: understanding what the generated code actually does, detecting logical flaws or unnecessary dependencies.
- Integrating security and compliance constraints (notably those related to the AI Act) from the development phase, not at the end of the chain.
The software and applications delivered in 2024 are not fundamentally different in their architecture. It is the way they are produced that is changing, with a layer of automation that accelerates certain tasks but creates new ones.
The IT trends of 2024 are not just a list of emerging technologies. The European regulatory framework redefines the conditions for deploying AI and cybersecurity. The energy constraint sets a physical limit to the race for ever-larger models. And assisted development tools redistribute the expected skills among developers. What distinguishes this year from previous ones is that these three pressures are exerting themselves simultaneously on the same teams.
