The high-tech landscape of 2024 is structured around three technical axes: the regulation of generative artificial intelligence, the migration of AI models to local devices, and the integration of security from the design phase of connected products. These high-tech trends are not a distant prospect; they are already altering purchasing decisions, software architectures, and compliance constraints for both businesses and individuals.
Embedded AI on smartphones and PCs: the end of the all-cloud
Most articles on the 2024 technology trends focus on the race for large generative models. However, the most structuring movement is at the opposite end: compact AI models designed to run locally, directly on a phone or laptop, without a permanent cloud connection.
Further reading : How to Access Your Emails: The Different Steps to Log in to Hotmail
Google has launched Gemini Nano, and Meta has released LLaMA 3 8B. These models fit within the RAM of a consumer device. Three technical reasons explain this shift: the reduction in inference cost (no server request), near-zero latency, and the preservation of personal data on the device.
To keep track of these hardware and software developments over the months, the high-tech site Info Geeks regularly covers benchmarks and new embedded models.
Recommended read : The essential guide to better understand the corporate decision-maker file in 2024
The benchmarks published in 2024 by MLCommons confirm a growing adoption of this approach. For common tasks (text summarization, photo sorting, response suggestions), a well-optimized compact model competes with a large cloud-hosted model, consuming a fraction of the energy.
What this changes for the consumer
A voice assistant that works in airplane mode. A photo app that sorts and edits without sending the pictures to a remote server. Embedded AI makes the user less dependent on their connection and on the privacy policy of a third party.
The trade-off is in power: complex tasks (long generation, analysis of large documents) remain more efficient in the cloud. The hybrid model, where the device handles simple requests and delegates heavier ones, is emerging as the dominant architecture.
European AI Act and compliance: the concrete regulatory constraint
The European AI Act was formally adopted in 2024. This regulation introduces a specific regime for general-purpose AI systems, a category that encompasses generative models used in chatbots, automated writing tools, and coding assistants.
The obligations are precise:
- Transparency regarding the training data used to build the model, with technical documentation accessible to regulatory authorities
- Management of illegal content generated by the system, with reporting and removal mechanisms
- Classification by risk level, with applications related to health, recruitment, or financial scoring being subject to the strictest constraints
These obligations will gradually come into effect starting in 2025. In France, the CNIL published several specific guidelines in 2024 on chatbots and generative AI, emphasizing the minimization of personal data and the prohibition of manipulative uses for sensitive profiling.
Direct impact on companies’ tech projects
The compliance budget becomes a distinct technical item. A company deploying a customer chatbot powered by a generative model must document the source of the training data, audit the responses produced, and provide a mechanism for human recourse.
The choice of AI provider is also changing. A model whose publisher does not provide documentation on the datasets used exposes the client company to a risk of non-compliance. This constraint favors transparent players and, by extension, open-source models whose architecture is publicly verifiable.
Data security and real-time threat management
The proliferation of connected devices and cloud services has mechanically expanded the attack surface. In 2024, the most tangible technological trend in security is the shift from a perimeter protection model (firewall around the network) to a continuous threat exposure management.
This concept, referred to by the acronym CTEM (Continuous Threat Exposure Management), relies on a continuous cycle:
- Real-time mapping of all exploitable entry points (web applications, APIs, mobile devices, cloud accounts)
- Simulation of automated attacks to identify vulnerabilities before an attacker can exploit them
- Prioritization of patches based on actual impact on the business, not just according to a theoretical severity score
This approach changes the role of IT teams. Security is no longer an annual audit but a daily process, integrated into development pipelines and product updates.
Connected gadgets and daily privacy
For the general public, this evolution translates into increased demands on connected gadgets: surveillance cameras, mixed reality headsets, sports watches. A device that no longer receives security patches becomes an intrusion vector into the home network.
Manufacturers that publish a guaranteed support duration (security updates for five years, for example) provide an objective selection criterion. This data, still rarely highlighted in product sheets, should become a standard for comparison alongside price or battery life.
The common thread of these 2024 tech trends can be summed up in one word: proximity. Artificial intelligence is migrating to local devices, regulation is bringing the responsibility closer to the deployer, and security is being integrated as close as possible to the code and hardware. The cycle of innovation is not slowing down, but it is getting closer to the ground.