A small British company, Cosine, is preparing to launch the UK’s first large language model (LLM) trained domestically, aiming to establish a homegrown alternative to dominant American and Chinese AI providers. Backed by partners including BAE Systems, Barclays, and Vodafone, Cosine plans to unveil its model, named Lumen Sovereign, this November.
The global AI landscape is predominantly shaped by the United States and China. US companies hold a lead in advanced capabilities, racing toward more human-like intelligence, while China is noted for rapid adoption, offering cheaper, more accessible AI systems. Against this backdrop, geopolitical tensions and policy decisions underscore the urgency for sovereign AI efforts. For instance, a temporary block imposed by former US President Donald Trump on foreign access to an Anthropic model earlier this year demonstrated how dependence on US technology can be disrupted unilaterally—even affecting allies.
In the UK, key sectors such as banking, defense, and telecommunications express concern about over-reliance on foreign AI providers. Political and security factors limit the use of Chinese models in regulated environments, while rising costs of US cloud-based AI services add further pressure. The financial industry watches closely as companies like OpenAI introduce AI-driven personal finance tools currently available only in the US, fearing potential customer losses if such innovations expand internationally.
Cosine, established in 2022, has secured $15 million in funding—significantly less than the billions invested in US giants like OpenAI—and employs around 30 staff. The company benefited from the UK government’s Sovereign AI initiative, which granted free access to the Isambard-AI supercomputer, valued at approximately $200 million, for training purposes. Cosine’s co-founder, Yang Li, highlights that their model is designed with a focused purpose rather than broad human-level intelligence. It targets specific regulated applications such as identity verification and HR compliance, distinguishing it from US frontier AI labs pursuing generalized AI.
Alongside Cosine, another UK firm, Aveni, is pursuing a niche strategy focused on financial services. With 100 employees spread between Edinburgh and London and £30 million raised, Aveni develops FinLLM using French Mistral models—choosing a European partner to maintain regional alignment. Aveni’s CEO, Joseph Twigg, argues that many industries do not require massive, resource-intensive AI models. Instead, smaller, industry-specific AI systems can deliver up to 25 times greater cost efficiency, operate locally, and reduce computational demands, strengthening the case for sovereign AI.
As these British startups take on much larger American and Chinese competitors, their focused, regulated-sector approaches may offer an alternative path for countries seeking to balance innovation with security and control. While the global AI market remains dominated by major powers, the emergence of homegrown players reflects a growing desire for domestic solutions amid shifting political and technological dynamics.