Liz, I agree with much of what you’ve written. Britain absolutely needs to back British AI, build sovereign capability and create an environment where innovative companies can thrive.
But I have to be honest about my experience.
For the past two years I’ve been trying to put a British AI capability in front of government. I’ve submitted proposals, met officials, engaged with departments and sought to engage with the Sovereign AI Fund.
The problem isn’t just funding.
I can’t even get through the door.
And when you do finally receive a response, it often comes months later. This week, after months of waiting, I received a letter thanking me for my correspondence, explaining the department’s objectives, wishing me every success and closing the case. That wasn’t an assessment of the technology; it was an administrative response. 00017003 - letter out.pdf
AI doesn’t wait. Technology doesn’t wait. Markets don’t wait. While government is still deciding who should evaluate a proposal, the rest of the world has already moved on.
You rightly highlight the risks of brain drain and losing our most promising AI companies. From where I sit, that isn’t an abstract policy concern. It’s the commercial reality founders face every day.
I want to build this company in Britain. I want the jobs, the intellectual property, the investment, the tax revenues and the long-term economic value to stay here.
But wanting to build in Britain and being able to build in Britain are not the same thing.
If founders with genuinely novel technology cannot get in front of the people capable of evaluating it, cannot access timely decisions and cannot secure the support needed to scale, while other governments are actively opening their doors, then we shouldn’t be surprised when those founders start looking elsewhere.
That isn’t a lack of patriotism. It’s the reality of building a technology company in one of the fastest-moving industries on the planet.
Britain doesn’t simply have a funding challenge. It has a discovery challenge, an engagement challenge and, above all, a speed challenge.
If we genuinely want sovereign AI capability, we have to become much better at finding our innovators, evaluating them quickly and giving them a reason to stay. Otherwise, we risk creating exactly what you describe: world-class British companies whose innovation, investment, jobs and future growth are realised somewhere else.
So let me ask you a genuine question.
If you were in my position two years trying to get a novel British AI capability recognised, unable to get through the door, while other countries are actively encouraging founders to build there would you tell me to stay, or would you tell me to go?
I genuinely want Britain to succeed. That’s why this frustrates me so much. We don’t lack talent. We don’t lack ambition. We are simply too slow. And in AI, being too slow is often the same as being too late.
NeuralMimicry strongly supports the central argument here: the next 24 months matter, and the UK should aim to shape the development of AI rather than simply consume systems designed, owned and governed elsewhere.
The emphasis on sovereign compute, British companies, government procurement and regional industrial capability is particularly welcome. AI sovereignty is not achieved through data centres alone; it requires UK-owned intellectual property, software architectures, skills, research, manufacturing routes and customers willing to adopt home-grown technology.
There is, however, an important point of nuance. Frontier large language models are a significant part of AI, but they are not the whole field. The dominant approach remains largely based on static machine-learning models: train, deploy, infer and periodically retrain. Scaling this approach has produced remarkable results, but it also brings substantial energy, data, hardware and dependency costs.
NeuralMimicry is exploring a complementary direction through AARNN: a distributed, self-evolving and continual-learning AI architecture inspired by neuromorphic principles. The aim is not to claim that neuromorphic or dynamic AI should replace every GPU workload or frontier model. Rather, different forms of intelligence should be applied where they are strongest.
Large models may remain valuable for language, knowledge access and broad reasoning. Dynamic, event-driven systems may be better suited to low-latency edge applications, robotics, sensing, cyber defence and environments where systems must continue adapting after deployment.
This also reinforces the article’s security argument. Offensive systems increasingly operate at machine speed. Defensive systems cannot rely solely on human reaction times or boundary controls. NeuralMimicry’s work includes an AI-swarm security agent intended to monitor, detect and respond to malicious behaviour, while preserving human authority over important decisions.
The article is right that the UK needs globally significant AI companies. It should also ensure that investment is not restricted to copying today’s leading architecture at smaller scale. Sovereignty means retaining the freedom to investigate alternative architectures that could define the next technological wave.
The UK has strong universities, engineers, industrial sectors and regional capability. What emerging companies often lack is not ambition or technical substance, but timely funding, compute access, commercial support and an early customer prepared to validate the technology.
NeuralMimicry is ready to contribute to that effort: building in Britain, retaining British ownership, and pursuing a form of AI intended to remain adaptive, efficient, governable and useful in the real world.
Secretary of State, this mandate is exactly right. The 24-month window is not an exaggeration; it is a strict operational reality.
As you rightly pointed out, securing British data—whether it is NHS records, defence intelligence, or regulated financial ledgers—is a matter of absolute national security. However, right now, the public sector and highly regulated industries are facing a severe "compliance lockout." They cannot safely deploy frontier AI if it means endlessly copying sovereign data into third-party, vendor-hosted cloud lakes.
At Open Portfolio, we are actively building the architectural solution to this exact bottleneck right here in Manchester. By deploying a stateless "Bring Your Own Cloud" (BYOC) boundary layer, we allow institutions to keep their data custody, identity providers, and approved storage completely intact, running AI inference safely at the edge.
This is the exact underlying plumbing required to make the Sovereign AI mandate a reality without expanding the state's cyber-risk footprint. We are fully aligned with the mission to build globally leading, privacy-preserving infrastructure outside of the South East, and we stand ready to supply the boundary architecture the public sector needs to adopt AI securely today.
Abba Lawal
Head of AI & Ecosystem Partnerships, Open Portfolio
Great article. Welcome to SS. Further reading. Enjoy
https://kaipability.substack.com/p/is-trl-dead-why-mcrl-might-be-what
https://kaipability.substack.com/p/the-uk-is-world-class-at-innovation
Liz, I agree with much of what you’ve written. Britain absolutely needs to back British AI, build sovereign capability and create an environment where innovative companies can thrive.
But I have to be honest about my experience.
For the past two years I’ve been trying to put a British AI capability in front of government. I’ve submitted proposals, met officials, engaged with departments and sought to engage with the Sovereign AI Fund.
The problem isn’t just funding.
I can’t even get through the door.
And when you do finally receive a response, it often comes months later. This week, after months of waiting, I received a letter thanking me for my correspondence, explaining the department’s objectives, wishing me every success and closing the case. That wasn’t an assessment of the technology; it was an administrative response. 00017003 - letter out.pdf
AI doesn’t wait. Technology doesn’t wait. Markets don’t wait. While government is still deciding who should evaluate a proposal, the rest of the world has already moved on.
You rightly highlight the risks of brain drain and losing our most promising AI companies. From where I sit, that isn’t an abstract policy concern. It’s the commercial reality founders face every day.
I want to build this company in Britain. I want the jobs, the intellectual property, the investment, the tax revenues and the long-term economic value to stay here.
But wanting to build in Britain and being able to build in Britain are not the same thing.
If founders with genuinely novel technology cannot get in front of the people capable of evaluating it, cannot access timely decisions and cannot secure the support needed to scale, while other governments are actively opening their doors, then we shouldn’t be surprised when those founders start looking elsewhere.
That isn’t a lack of patriotism. It’s the reality of building a technology company in one of the fastest-moving industries on the planet.
Britain doesn’t simply have a funding challenge. It has a discovery challenge, an engagement challenge and, above all, a speed challenge.
If we genuinely want sovereign AI capability, we have to become much better at finding our innovators, evaluating them quickly and giving them a reason to stay. Otherwise, we risk creating exactly what you describe: world-class British companies whose innovation, investment, jobs and future growth are realised somewhere else.
So let me ask you a genuine question.
If you were in my position two years trying to get a novel British AI capability recognised, unable to get through the door, while other countries are actively encouraging founders to build there would you tell me to stay, or would you tell me to go?
I genuinely want Britain to succeed. That’s why this frustrates me so much. We don’t lack talent. We don’t lack ambition. We are simply too slow. And in AI, being too slow is often the same as being too late.
NeuralMimicry strongly supports the central argument here: the next 24 months matter, and the UK should aim to shape the development of AI rather than simply consume systems designed, owned and governed elsewhere.
The emphasis on sovereign compute, British companies, government procurement and regional industrial capability is particularly welcome. AI sovereignty is not achieved through data centres alone; it requires UK-owned intellectual property, software architectures, skills, research, manufacturing routes and customers willing to adopt home-grown technology.
There is, however, an important point of nuance. Frontier large language models are a significant part of AI, but they are not the whole field. The dominant approach remains largely based on static machine-learning models: train, deploy, infer and periodically retrain. Scaling this approach has produced remarkable results, but it also brings substantial energy, data, hardware and dependency costs.
NeuralMimicry is exploring a complementary direction through AARNN: a distributed, self-evolving and continual-learning AI architecture inspired by neuromorphic principles. The aim is not to claim that neuromorphic or dynamic AI should replace every GPU workload or frontier model. Rather, different forms of intelligence should be applied where they are strongest.
Large models may remain valuable for language, knowledge access and broad reasoning. Dynamic, event-driven systems may be better suited to low-latency edge applications, robotics, sensing, cyber defence and environments where systems must continue adapting after deployment.
This also reinforces the article’s security argument. Offensive systems increasingly operate at machine speed. Defensive systems cannot rely solely on human reaction times or boundary controls. NeuralMimicry’s work includes an AI-swarm security agent intended to monitor, detect and respond to malicious behaviour, while preserving human authority over important decisions.
The article is right that the UK needs globally significant AI companies. It should also ensure that investment is not restricted to copying today’s leading architecture at smaller scale. Sovereignty means retaining the freedom to investigate alternative architectures that could define the next technological wave.
The UK has strong universities, engineers, industrial sectors and regional capability. What emerging companies often lack is not ambition or technical substance, but timely funding, compute access, commercial support and an early customer prepared to validate the technology.
NeuralMimicry is ready to contribute to that effort: building in Britain, retaining British ownership, and pursuing a form of AI intended to remain adaptive, efficient, governable and useful in the real world.
https://github.com/neuralmimicry
Nope. Breathless bandwagon leaping leads all to the same hole.
Secretary of State, this mandate is exactly right. The 24-month window is not an exaggeration; it is a strict operational reality.
As you rightly pointed out, securing British data—whether it is NHS records, defence intelligence, or regulated financial ledgers—is a matter of absolute national security. However, right now, the public sector and highly regulated industries are facing a severe "compliance lockout." They cannot safely deploy frontier AI if it means endlessly copying sovereign data into third-party, vendor-hosted cloud lakes.
At Open Portfolio, we are actively building the architectural solution to this exact bottleneck right here in Manchester. By deploying a stateless "Bring Your Own Cloud" (BYOC) boundary layer, we allow institutions to keep their data custody, identity providers, and approved storage completely intact, running AI inference safely at the edge.
This is the exact underlying plumbing required to make the Sovereign AI mandate a reality without expanding the state's cyber-risk footprint. We are fully aligned with the mission to build globally leading, privacy-preserving infrastructure outside of the South East, and we stand ready to supply the boundary architecture the public sector needs to adopt AI securely today.
Abba Lawal
Head of AI & Ecosystem Partnerships, Open Portfolio
https://www.openportfolio.co.uk/