Key Takeaways
- Portugal launched Amalia, an open-source large language model funded with €5.5 million in EU recovery funds
- The model targets public institutions and businesses, not direct consumer use, with initial deployments in museums, the Navy, and education
- Portugal joins France (Mistral AI) and Germany (Aleph Alpha) in Europe's coordinated push to reduce dependence on US AI providers
Portugal unveiled Amalia on Wednesday, the country's first open-source AI model, backed by €5.5 million ($6.26 million) in EU recovery funds. Named after legendary fado singer Amália Rodrigues, the large language model is designed as foundation technology for Portuguese institutions rather than a consumer chatbot.
The launch slots Portugal into a growing European effort to build alternatives to American AI dominance. France has Mistral AI. Germany backs Aleph Alpha. Now Portugal has Amalia. Each represents the same bet: that depending entirely on OpenAI, Google, and Anthropic for AI infrastructure carries strategic risk.
What can Amalia actually do?
Unlike ChatGPT or Claude, Amalia is not meant for public interaction. It is a base model that public institutions, companies, universities, and researchers can adapt for specific applications. Think of it as AI infrastructure rather than an AI product.
Four initial applications are already in development: a virtual guide for Portugal's museums, decision-support tools for the Portuguese Navy, an AI-powered teaching assistant for lesson planning, and a digital assistant for delivering public services to citizens.
The model, its training dataset, and source code are all released under an open-source license. This means any Portuguese organization can inspect, modify, and deploy it without licensing fees or vendor lock-in.
Why European governments are building their own AI
Prime Minister Luís Montenegro framed the launch in explicit sovereignty terms. "Europe's strategic autonomy is today, perhaps more than ever, tied to AI," he said at the event. "This model will enable us to face the coming decades with greater sovereignty and less dependence."
The concern is not abstract. European governments and enterprises currently route sensitive data through American AI providers, subject to US jurisdiction and corporate priorities. Building local models gives institutions an alternative where data stays within European legal frameworks.
Montenegro promised continued investment: "We will continue to invest heavily in this project." He cited banking, insurance, telecommunications, and industrial applications as sectors where Amalia could boost productivity while ensuring security.
The computing muscle behind Amalia
Training large language models requires serious compute. Portugal addressed this by tapping into two supercomputers: Deucalion (hosted in Portugal) and MareNostrum 5 (Barcelona). This high-performance computing access distinguishes Amalia from efforts that might otherwise stall on infrastructure costs.
A consortium of Portuguese universities and research institutions developed the model with government backing. The €5.5 million came from EU recovery funds, part of the bloc's broader push to build sovereign tech capacity after COVID-19 exposed supply chain vulnerabilities.
How this fits Europe's AI strategy
Portugal's move is not isolated. The EU has allocated over €1 billion to AI development through the AI Act and Digital Europe Programme. France's Mistral AI raised over $1 billion and now offers models competitive with OpenAI's offerings. Germany's Aleph Alpha has focused on enterprise and government contracts.
The common thread: open-source or European-controlled models that reduce dependence on Silicon Valley. Whether this strategy produces globally competitive AI remains an open question. But for government applications where data sovereignty matters more than bleeding-edge capabilities, homegrown models make sense.
Logicity's Take
Portugal's €5.5 million investment is modest compared to the billions flowing into OpenAI or Anthropic. Amalia will not compete with GPT-4 on benchmarks. That is not the point. For CTOs at European enterprises handling regulated data, or governments wary of US cloud dependencies, a functional open-source model with clear provenance solves a real problem. The question is whether European institutions will actually adopt these sovereign alternatives or default to the more capable American options when procurement decisions come down to features.
Frequently Asked Questions
Is Amalia available for public use like ChatGPT?
No. Amalia is a foundation model intended for institutions, companies, and researchers to build applications on top of, not a consumer-facing chatbot.
What does open-source mean for Amalia?
The model weights, training dataset, and source code are all released under an open-source license. Organizations can inspect, modify, and deploy the model without licensing fees.
How does Amalia compare to models from OpenAI or Anthropic?
Details on model size and benchmarks have not been released, but the focus is on European language support and data sovereignty rather than competing on raw capability with well-funded US models.
Which other European countries have launched sovereign AI models?
France backs Mistral AI, and Germany supports Aleph Alpha. Both aim to provide European alternatives to US-developed AI systems.
What applications is Amalia being used for initially?
A museum virtual guide, Portuguese Navy decision-support tools, an AI teaching assistant, and a digital assistant for public services.
How Anthropic is addressing enterprise AI deployment concerns
Need Help Implementing This?
If you are evaluating open-source AI models for your organization, or need help setting up infrastructure for local LLM deployment, reach out to Logicity's consulting team for architecture guidance.
Source: Tech-Economic Times / ET
Manaal Khan
Tech & Innovation Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
