Today, we published an Open Letter on sovereign AI for the United Kingdom, alongside a coalition of leaders across critical national infrastructure. AI is a question of national capability. While Britain possesses world-class research and software talent, we risk becoming a customer in this technological shift rather than a serious builder. Countries with reliable access to compute and strategic models will have greater resilience. Those that lack these capabilities risk depending on infrastructure and decisions determined elsewhere. Sovereign AI means the ability to govern, secure, evaluate, and sustain strategically important AI capabilities under UK jurisdiction. We must ensure high-consequence systems can be inspected and governed domestically to strengthen public confidence and national resilience, while remaining engaged internationally. We are urging the Government to treat AI as strategic national infrastructure. We need a long-term plan that builds UK capability in critical domains, uses public procurement to create demand, and aligns investment with clear national priorities. The coming year will shape Britain’s position for decades to come. You can read the full letter attached and explore our recommendations in more detail on our open letter page here: cosine.sh/sovereign-ai-open-… @andyburnham @KanishkaNarayan

We are pleased to welcome @edfenergy to the Lumen Sovereign coalition. EDF brings essential insight from the highly regulated energy sector. Niels Roberts, Chief Information Officer at EDF (UK), shared the following: “AI is already in use across the UK’s energy system. But for it to be widely used in highly regulated critical national infrastructure, we need to ensure that it is both reliable and resilient. Sovereignty will be crucial. EDF is therefore delighted to bring its unique perspective across nuclear, renewables and energy supply to the development of the UK’s own sovereign AI capability.” Together, we are ensuring that Lumen Sovereign is shaped around the operational realities and assurance requirements of complex, high-stakes environments.

Sovereign AI dominates global conversations, yet most discussions focus on policy rather than the engineering realities of building these systems. Today, we’re publishing a practitioner’s guide to sovereign AI, written specifically for teams worldwide designing and implementing these architectures. At its core is our AI Value Chain framework: a layer-by-layer technical breakdown of the dependencies at every stage: → Hardware → Infrastructure → Data → Models → Serving & Orchestration → Application We also explore a UK case study, featuring our work engineering Lumen Sovereign, Britain’s first fully sovereign frontier model. This was a significant collaborative effort. Thank you to our Cosine contributors (Niall Devlin, Robert Gibson) and our external collaborators (Harshali Ranjan, Kristoffer Bjärkefur, and Luis Eduardo San Martin) for sharing their expertise. cosine.sh/blog/building-sove…

We’re pleased to welcome Sopra Steria to the Lumen Sovereign coalition. They bring deep expertise in delivering critical national infrastructure. Craig Wilson, CEO of Sopra Steria UK, said: “Sovereign AI is moving from policy debate to operational reality. For organisations delivering critical services, the opportunity is to bring people, technology and transformation together to improve productivity, while maintaining trust and control. This depends on how AI is built, deployed and governed. By joining Cosine’s coalition, we will bring our experience of delivering complex, regulated services at scale, to help turn sovereign AI into practical solutions that create value safely and securely.” Together, we’re accelerating safe, sovereign AI adoption across highly regulated UK sectors.

If you’re as concerned about AI sovereignty as I am, I recently joined Ben Ansell on BBC Radio 4. We got into the realities of building sovereign technology in the UK and what we’re building at @CosineAI. @benwansell bbc.co.uk/sounds/play/m00318…

Independent assurance and safety are core to how we design and govern our model. We’re pleased to welcome Advai to the Lumen Sovereign coalition. Advai shared the following: “We are excited to be collaborating with Cosine on the Lumen Sovereign model build, bringing Advai’s extensive experience in testing, red teaming and monitoring AI models and systems to the coalition. If sovereign AI is going to be trusted in high-assurance and regulated environments, safety, security and accountability need to shape how those models are designed and built from the ground-up. Doing so creates a foundation for effective, reliable and trustworthy models that can be deployed at scale.” Embedding independent testing throughout development strengthens the foundations Lumen Sovereign needs to operate safely where trust matters most.

We’re excited to collaborate with Llywodraeth Cymru / Welsh Government. 🏴󠁧󠁢󠁷󠁬󠁳󠁿 We are bringing Welsh language data into the training pipeline for our Lumen Sovereign model, ensuring our sovereign AI reflects the diverse languages and culture of the UK. More details soon! - Rydym yn gyffrous i gydweithio â Llywodraeth Cymru. 🏴󠁧󠁢󠁷󠁬󠁳󠁿 Rydym yn cyflwyno data yn yr iaith Gymraeg i'r broses hyfforddi ar gyfer ein model Lumen Sovereign, gan sicrhau bod ein modelau AI sofran yn adlewyrchu ieithoedd a diwylliant amrywiol y DU. Edrychwn ymlaen at rannu mwy o fanylion yn fuan! @WelshGovernment

We’re proud to welcome Kainos to the Lumen Sovereign coalition. Building frontier AI for UK-regulated environments means navigating some of the world’s most rigorous compliance and security landscapes. Kainos brings crucial frontline expertise to our work. In the coming months, we will be collaborating to define clear, practical pathways for adopting sovereign AI across the UK's regulated environments.

Teams want greater control over how agents are directed and evaluated. Plausible solutions aren’t enough. Agents must work against explicit constraints and verify results. We’re building our harness to do exactly that. With Goals, define objectives and criteria – get evidence-backed results.

What does "Sovereign AI" actually mean for regulated industries? Join @CoreWeave and @CosineAI on September 10th to learn how to deploy frontier AI securely and how to prove it's compliant. In a live demo, you'll see how to use W&B Weave to trace every compliance check and red-team probe, helping you build reproducible, evidence-backed approval records instead of relying on self-attestations Register here: utm.io/usD9p

An early version of Lumen Sovereign – our AI model trained entirely in the UK – is running locally on an Apple Mac Studio in our office. This 200B baseline is just the beginning. We're now breaking ground on a 1 trillion-parameter model, actively synthesising RL data from real-world use cases for complex tasks provided by our coalition and public RFI. 📰 Keep up to date with Lumen Sovereign by joining our newsletter: tally.so/r/1AzGq1

PA Consulting is joining the Lumen Sovereign Coalition. They bring deep domain expertise, helping shape practical deployments of sovereign AI across regulated sectors. We’re also collaborating on a nuclear-specific AI model for civil and defence sectors, building on PA’s and Jacobs’ work with Sellafield and Nuclear Decommissioning Authority. Together, we’re making sovereign AI a practical, trusted capability for critical national infrastructure.

What does sovereign AI actually mean for regulated industries? Join Cosine’s Tom Dowley alongside Weights & Biases by CoreWeave on 10 September to discuss practical sovereignty, infrastructure security, and our Lumen Sovereign deployment. Register now: wandb.ai/site/resources/even…

"We've experimented with other players, but we always come back to Nebius." — @CosineAI, on scaling AI for regulated industries in fully air-gapped environments. 🎥

Today, we are proud to share that EY has joined the Lumen Sovereign coalition. EY brings deep industry knowledge and expertise in financial services use cases, AI governance, and risk management. This expertise will help ensure the UK’s sovereign AI capability meets the demands of critical financial workflows and other highly regulated environments. Together, Cosine and EY will help shape a trusted, resilient, and independent AI capability that gives UK institutions greater choice and confidence in how they adopt and deploy AI.