The latest developments in artificial intelligence are raising an increasingly important question: as AI becomes more accessible, is power over the technology really becoming more distributed?
Over the past week, developments involving open-weight AI models, AI infrastructure, accessibility and governance have shown how complicated the answer may be.
Open-Weight AI and the Question of Control
Meta introduced Muse Glimmer, a new open-weight AI model designed to handle smaller agentic tasks locally, including on a single GPU.
Mark Zuckerberg also argued that concentrating advanced AI capabilities among a small number of companies or institutions could itself create risks.
But another view emerged from Washington.
US Senator Bernie Sanders called on Meta, OpenAI and Anthropic to pause the development of advanced AI systems, highlighting concerns around increasingly autonomous AI.
Two very different responses to the same acceleration are emerging: open AI further, or slow its development down.
The $500 Billion AI Infrastructure Question
While models become more accessible, another AI race is happening underneath them.
Nvidia is working with major Wall Street institutions on financing that could support up to $500 billion in AI infrastructure.
AI models may become cheaper and open-weight systems more capable, but the chips, data centres, energy and capital required to power them remain enormously expensive.
If access to models becomes distributed while the infrastructure remains concentrated, how democratic is AI really becoming?
AI Accessibility in Ireland: Vision Ireland
Ireland provided a tangible example of what wider AI access can look like.
Meta announced plans to donate 15,000 Ray-Ban Meta smart glasses to Vision Ireland, alongside training and support for people living with sight loss.
It is an important reminder that AI accessibility is not only about making models open or affordable.
It can also mean using AI to create new possibilities in everyday life.
What Could This Mean for AI in Ireland?
The question of access is arriving alongside another major transition for AI in Ireland.
As the EU AI Act moves into implementation, the AI Office of Ireland is taking a central role in the country’s emerging AI governance framework.
Ireland may not build the world’s largest AI model or compete directly with the enormous infrastructure investments being made elsewhere.
But it has another opportunity: becoming a place that understands how AI can be adopted, governed and applied responsibly.
Perhaps creating value in the AI era will not only be about building the most powerful model.
It will also be about who can access AI, who controls its infrastructure, who sets its rules and, ultimately, who benefits from it.
As AI becomes more open, is power really becoming more distributed, or is that power simply moving somewhere else?
