What indicators allow us to measure the most significant technological changes of 2025-2026? Between the widespread adoption of dedicated artificial intelligence processors in consumer PCs, the gradual implementation of the European AI regulation, and breakthroughs in autonomous robotics, the tech sector is undergoing a phase where regulation catches up with hardware innovation. Three major axes are emerging, each with very concrete implications for businesses and individuals.
Embedded NPUs and Copilot+ PC: Mapping the Forces at Play
The shift towards PCs capable of running AI models locally, without relying on the cloud, is redefining the computing market. Microsoft officially announced this direction with its Copilot+ PC range unveiled in May 2024, followed by major manufacturers.
Read also : The latest news not to be missed: follow the news continuously every day
| Manufacturer | Chip / NPU | Main Use |
|---|---|---|
| Intel | Meteor Lake, then Lunar Lake | Consumer and professional laptops |
| Lenovo | Integration of Intel and Qualcomm NPUs | ThinkPad and Yoga range |
| Dell | Intel and Snapdragon X NPU | XPS, Latitude for business |
| HP | Intel and Qualcomm NPU | Spectre, EliteBook |
| Asus | Intel and Qualcomm NPU | ZenBook, ProArt for creatives |
The common feature of these machines: a Neural Processing Unit dedicated to embedded AI, capable of running generative assistants (text, image, translation) directly on the device. Latency decreases, data privacy improves, and the dependence on a constant connection disappears.
The news from SOS Technologies highlights how this transition to local AI is also changing the expectations of IT services in companies, which must rethink their machine fleets.
See also : Discover the latest news and information about the independent movement in France
Microsoft has also introduced Windows Recall, a feature that records and indexes user activity thanks to the NPU. This announcement has raised privacy concerns that directly relate to the European regulatory framework.
European AI Act: What the Regulation Changes for Tech Companies
The regulation (EU) 2024/1689 on artificial intelligence was published in the Official Journal of the EU on July 12, 2024. Its implementation will be rolled out gradually between 2025 and 2026, depending on the risk level of the concerned systems.
Two categories of constraints deserve particular attention.
Transparency Obligations for Generative AI
Providers of generative models must now document the training data used and label the content produced by AI. This requirement affects both American giants (OpenAI, Google, Meta) and European players like Mistral AI.
For French companies integrating these models into their digital products or services, compliance with the labeling of generated content becomes a legal obligation, not an ethical choice.
High-Risk Systems: Health, Finance, Recruitment
AI systems classified as high risk (medical diagnosis, financial scoring, application sorting) will have to meet strict requirements for documentation, auditing, and human oversight. The penalties provided for in the text are proportional to the global revenue of the company.
- Automated recruitment tools must provide a readable explanation of their selection criteria to candidates and employers.
- Medical devices incorporating generative AI will be subject to dual certification: that of the AI regulation and that of existing medical devices.
- Bank scoring systems must allow for human contestation of each automated decision.
In contrast, systems considered to be minimal risk (photo filters, basic customer service chatbots) are only subject to limited transparency obligations.
Autonomous Robotics and Drones: Field Advances of 2026
Robotics is experiencing an acceleration driven by two factors: the decreasing cost of sensors and the ability of AI models to train in virtual environments before physical deployment. Nvidia has introduced Cosmos, its platform that allows robots to learn in realistic simulations before interacting with the real world.
This principle of virtual training followed by field transfer significantly reduces development time. A logistics robot can thus accumulate months of simulated experience in just a few days of computation.
On the drone side, the year 2026 marks a turning point in professional uses. Logistics and transport are adopting drones for last-mile delivery and infrastructure inspection. Valeo, for example, is showcasing innovations related to autonomous driving and embedded sensors for the Chinese market at Auto China 2026.
AI in Business: Measurable Productivity and Data Governance
The adoption of artificial intelligence in business is no longer limited to experiments. Orange has documented the productivity gains achieved through AI in its own operations, while emphasizing the need for strict governance of the data used to train the models.
Data governance for training data has become the main barrier to adoption, more so than the cost of tools. French companies that have not mapped their internal data struggle to deploy generative AI solutions suited to their businesses.
- IT services must audit datasets before any training to avoid biases and GDPR violations.
- Legal departments are now involved at the outset of AI projects, rather than just at final validation.
- Business teams participate in defining use cases, reducing the gap between technological promises and real needs.
The tech sector in 2025-2026 is characterized by a convergence of local computing power, a binding European regulatory framework, and maturity of use cases in business. PC manufacturers are integrating AI at the hardware level, Europe is imposing transparency rules on generative models, and organizations are finally structuring their data governance. The next step will largely depend on companies’ ability to turn these regulatory constraints into a competitive advantage in the digital market.