The race to build a national language model is no longer a Silicon Valley obsession. It has become a matter of economic survival. From Delhi to Berlin, governments are pouring billions into creating what they call 'sovereign AI' large language models trained on local data, in local languages, reflecting local values.
For the workers I speak with in Rotherham and Rochdale, this might sound like abstract techno-babble. But the consequences are already landing on kitchen tables. When a government runs its welfare chatbot on an American-owned model, the algorithm doesn't understand why a steelworker from Scunthorpe might type 'redundant' differently to a banker in Canary Wharf. It doesn't grasp the regional dialects, the local benefit rules, or the dignity of a manual worker navigating a digital form.
The push for sovereign LLMs is a push for control. The European Union is funding 'OpenGPT' projects. India is building 'BharatGPT' in 22 languages. France has 'Mistral'. The UK, slow off the mark, is now scrambling with the 'Foundation Model Taskforce'. The logic is simple: if AI is the new electricity, no country wants to be dependent on a foreign power plant that might switch off the grid.
But there is a cost. These models require vast computing power, rare minerals for chips, and energy that could heat homes. In a time of rising bills, is this the right priority? Unions worry that national AI projects will be used to automate public sector jobs without consultation. The PCS union has already warned that 'digital sovereignty' could become a cover for 'digital cuts'.
Yet the alternative is worse. Without a sovereign LLM, Britain's public services will be shaped by the commercial interests of Silicon Valley. An AI trained on American consumer data will not understand the nuances of the NHS, the complexities of Universal Credit, or the history of a mining community. It will recommend policies based on profit, not people.
The question is not whether we need sovereign AI. The question is: who will build it and who will benefit? If it is built by private contractors, behind closed doors, it will be another tool for inequality. If it is built in the open, with trade unions, civil society, and workers at the table, it could be a tool for public good.
This is a fight for the real economy. The digital economy is not separate. It is the same economy, just running on different machines. And if we get it wrong, the price will be paid in lost jobs, eroded public services, and a further hollowing out of the regions.
Every nation needs a LLM. But every nation also needs a plan for who gets left behind.
