Meet Amalia: An AI with a Portuguese Soul
On July 1, 2026, Portugal officially launched Amalia, its first large-scale language model (LLM) designed specifically for the Portuguese language and its cultural nuances. The name is a tribute to Amália Rodrigues, the iconic fado singer whose voice
is deeply connected to the nation's identity. The name is also an acronym for Automatic Multimodal Language Assistant with Artificial Intelligence. Developed by a consortium of Portuguese universities and research institutions with €5.5 million in government and EU funding, Amalia is not intended to be a direct consumer-facing rival to services like ChatGPT. Instead, it is a foundational, open-source platform. This means its model, training data, and source code are freely available for companies, researchers, and public institutions to build upon.
More Than Code: A Cultural & Linguistic Shield
A primary motivation behind Amalia is cultural and linguistic preservation. Global commercial AI models are overwhelmingly trained on data from the internet, which heavily favors American English and, in the case of Portuguese, the Brazilian variant. This often results in AI systems that flatten the unique grammar, idioms, and cultural context of European Portuguese. Amalia was trained specifically on Portuguese-language data to understand and generate text grounded in the nation's reality. This makes it a crucial tool for public services that need to communicate with citizens in a familiar and accurate register. By creating a model that truly speaks the nation's language, Portugal aims to ensure its digital future reflects its distinct cultural identity.
The Strategic Quest for AI Sovereignty
The Amalia project is a cornerstone of Portugal's broader National Artificial Intelligence Agenda, a strategy aimed at reducing technological dependence on the United States and China. The concept of “AI sovereignty” has gained traction across Europe as nations worry about foreign control over critical digital infrastructure. By developing its own foundational model, Portugal aims to ensure that sensitive data, especially within public administration, does not have to leave its national territory for processing by foreign-owned AI. Prime Minister Luís Montenegro described the initiative as essential for facing the coming decades with "greater sovereignty and less dependence." The open-source nature of Amalia is key to this strategy, designed to foster a domestic ecosystem of AI innovation.
A Growing European Trend
Portugal is not alone in this endeavor. Its move is part of a wider European push for digital autonomy. Countries like France and Germany have backed homegrown companies like Mistral AI and Aleph Alpha to provide credible alternatives to dominant U.S. models. Spain, Poland, and Denmark are also developing their own national LLMs. These efforts are complemented by pan-European projects like OpenEuroLLM and EUROPA, which aim to create multilingual models covering all official EU languages, reflecting a collective desire to compete on the global stage while upholding European values and regulations. These initiatives highlight a strategic shift, where developing proprietary AI is increasingly seen as a matter of national and regional security.
Ambition Meets Reality: The Road Ahead
Despite the strategic importance, the path forward has its challenges. The biggest hurdle is not just creating the model, but ensuring its widespread adoption. Getting universities, private companies, and government agencies to actively build applications on top of Amalia will be critical for its long-term success. The government has already outlined several initial applications, including an AI teaching assistant, a virtual guide for museums, and decision-support tools for the Portuguese Navy. By leveraging national supercomputing resources and fostering collaboration between academia and industry, Portugal hopes to turn its ambitious vision of AI sovereignty into a tangible economic and social reality.
