As artificial intelligence moves beyond computers and into vehicles, medical devices, factories, and critical infrastructure, European technology organizations have released a free reference guide designed to help engineers build safer, more reliable AI systems.
WHAT’S HAPPENING
Artificial intelligence is moving into the physical world, and European technology organizations are addressing a challenge that could affect how safely and efficiently these systems are developed: the lack of a shared technical language.
On October 15, 2026, Embedded France and six partner organizations, including the Eclipse Foundation, released the English-language version of the Structured Glossary of Embedded AI.
The publication contains more than 300 technical definitions organized into 13 chapters, covering everything from AI models and hardware architecture to cybersecurity, safety testing, regulatory compliance, and emerging technologies.
Originally published in French in January 2026, the glossary is now freely available in both languages.
The initiative brings together specialists from Embedded France, the Eclipse Foundation, CEA, Leminnov, Quantyss, the European Trustworthy AI Association, and Lianeo.
Their objective is straightforward: establish consistent terminology that engineers, researchers, software developers, manufacturers, and safety specialists can use when developing AI-powered physical systems.
WHY IT MATTERS
Most people encounter artificial intelligence through websites, smartphones, and conversational assistants.
But AI is increasingly being integrated directly into machines.
This technology, known as embedded AI, can operate inside vehicles, industrial robots, medical equipment, aircraft systems, and electrical infrastructure.
Unlike conventional cloud-based AI, embedded systems frequently operate with limited processing power, restricted memory, strict response-time requirements, and significant safety considerations.
A misunderstanding between engineering teams can create problems extending far beyond software performance.
For example, a machine-learning developer and a safety engineer might interpret terms such as reliability, validation, or real-time processing differently.
Those differences can lead to incorrect technical specifications, repeated testing, additional development expenses, or delayed product approvals.
The new glossary addresses this problem by establishing shared definitions across disciplines.
It also covers emerging technologies such as generative AI, agentic AI, and physical AI, helping organizations understand how these concepts fit into real-world systems.
WHO BENEFITS
Manufacturers: Automotive, aerospace, industrial, and medical-device companies can use consistent terminology when coordinating complex engineering projects.
AI developers: Software teams gain a common reference when collaborating with hardware engineers, cybersecurity specialists, and safety professionals.
Educational institutions: Universities and technical training programs can incorporate the glossary into courses focused on embedded AI and intelligent machines.
Smaller technology companies: Startups and businesses without extensive internal technical documentation gain access to a freely available reference developed by industry specialists.
Regulatory and compliance teams: Shared definitions can improve communication when preparing technical documentation and evaluating applicable standards.
The glossary is published under a Creative Commons Attribution 4.0 license, allowing organizations to reuse, translate, and adapt its contents, including commercially, provided appropriate attribution is given.
WHO LOSES
The initiative does not create an immediate commercial disadvantage for any particular industry.
However, organizations that continue relying on inconsistent terminology may face avoidable challenges.
Companies using poorly defined technical requirements: Ambiguous specifications can contribute to development delays, redesigns, and unnecessary expenses.
Engineering teams working in isolation: Departments that lack a common understanding of AI terminology may struggle to coordinate projects involving multiple technologies.
Businesses unprepared for AI compliance: Organizations developing safety-sensitive products may face additional difficulties documenting and validating systems.
Importantly, the glossary is not a mandatory international standard or certification program. Using it does not automatically establish that a product is safe or legally compliant.
WHAT HAPPENS NEXT
The glossary will remain an open, evolving resource hosted through the Eclipse Foundation’s Eclipse Aidge project.
Developers, researchers, and organizations can propose corrections, clarifications, and additional terminology through its public repository.
The publication is also scheduled for presentation at the January 26, 2027, Assises de l’Embarqué conference, which will focus on physical AI.
As intelligent machines become more sophisticated, communication between AI developers, hardware manufacturers, and safety specialists will become increasingly important.
Businesses developing embedded AI can begin using the glossary immediately as a reference when preparing specifications, coordinating teams, or building employee training materials.
The resource is available through the Eclipse Foundation’s Structured Glossary of Embedded AI.
The larger signal: Artificial intelligence is entering a new stage in which its success depends on more than powerful algorithms. As AI becomes part of machines that interact with people and the physical environment, consistent technical definitions, coordinated engineering practices, and clear safety requirements become essential. Europe’s open glossary represents a practical step toward building that foundation.
