British banks, telecoms, and arms makers have a new shared concern: that the AI they continue to work on is built, owned, and controlled in the United States. A startup less than three years old is betting they’ll pay to fix it.
Cosine, the UK’s frontier AI lab, has assembled a consortium of British institutions including Lumen Sovereign, which it bills as Britain’s first frontier AI model.
The signatories read like a call to the forefront of the country’s economy: BT, HSBC, Lloyds, NatWest, BAE Systems, Babcock, LSEG, PwC, Thales UK, Leonardo UK, and Telefónica Tech, each signing a memorandum of understanding to help shape the model’s use cases, security requirements, and governance.
The unveiling coincided with the opening of London Tech Week, where Prime Minister Keir Starmer launched an AI intervention strategy and announced nearly £400mn of new spending on specialist AI chips to increase the country’s computing capacity. Britain’s next AI champions, he said, should “start here, scale here and stay here.”
💜 for EU tech
The latest talk from the EU tech scene, a story from our genius founder Boris, and some incredible AI art. It’s free, every week, in your inbox. Register now!
The selling point is control. Lumen Sovereign will be fully trained on Isambard-AI, an Nvidia-powered Bristol supercomputer that is among the most powerful in Europe, using a computer awarded under the government’s £500mn Sovereign AI programme, which named Cosine in its first team in April.
Cosine says the model will not rely on foreign infrastructure at any stage, and can be installed in the customer’s own systems, including air-gap areas without connecting to external networks.
That’s the essence of the pitch.
For many companies, choosing an AI model is a purchasing decision; for a defense contractor, a bank running anti-money laundering checks, or an operator of critical infrastructure, legality and security come first. Sending classified systems, AML alerts, or clinical data to a server in a US data center is generally prohibited.
“Businesses are increasingly waking up to the risk of relying entirely on external suppliers,” said Cosine founder and CEO Alistair Pullen, who says closing in on vendors brings “security risk, dependency risk, and cost escalation risks.”
Cosine is an unusual candidate for the job, and an honest one.
Founded in 2022 by Pullen, Yang Li, and Sam Stenner, the Y Combinator-backed lab has raised just $8mn from investors including Lakestar, yet its coding models have topped independent benchmarks against OpenAI, Anthropic, Mistral, and DeepSeek for two years running. It supports more than 38 programming languages, including COBOL, Fortran, and Ada that still use British security systems and financial plumbing, and that many AI tools handle badly.
Lumen Sovereign, it says, will be built from scratch on a proprietary dataset spanning more than 30 managed workflows, rather than fine-tuned from an open-source model, with a target of late 2026.
Our alliance partners do not hide why they entered. “Cosine has given us a path to a completely UK-native and highly customized AI stack,” said Peter Passaro, director of AI and data at Babcock, pointing to the “highly complex security environments we work in.”
Key uses include cybersecurity, KYC and AML investigations, legal document review, and healthcare management, areas where UK AI adoption has stalled in these security concerns.
Cosine’s effort is the sharp end of Britain’s broader argument for AI autonomy, which is increasingly clashing with a harsh reality. The government has invested in home computing and chip supply, and is making private infrastructure deals with Nscale and Nvidia, even as prominent US projects such as OpenAI’s Stargate UK have stumbled.
But “independent” AI in Britain still likes to use American chips, Isambard-AI among them.
For now, these are memorandums, not contracts, and Lumen Sovereign is yet to be built. Training a border model from scratch to the certification standards required by defense and finance, and doing it by the end of the year, is a huge undertaking, and Britain has previously announced AI-sovereign ambitions.
The exception this time is the consortium: a chain of institutions that treat AI sovereignty not as a slogan but as a procurement requirement.
