This has been a busy week for AI policy thinking in the UK. The UK Foreign Secretary Yvette Cooper published an essay “Britain’s Place in the New World Order” to deliver a stark assessment of a more dangerous global era, the pace of technological change and how to tackle the AI safety question. In this, she argues that:

We can only take advantage of the amazing opportunities frontier technologies can bring if there is sufficient international consensus on safety arrangements and guardrails.

 On the same day the Mills Review by the Financial Conduct Authority (FCA) set out a roadmap for how UK industry regulators and government should respond to AI. 

Into this febrile environment, and with perfect timing, AI platform provider Kore.AI today announces a partnership with Atos to deliver Sovereign Agentic AI for UK public sector, financial services, healthcare, defence and critical national infrastructure.

We cannot wait for an AI equivalent of Hiroshima

Cooper’s essay claims the UK is “the leading voice on AI security.” This is a moot point, as globally, the EU are racing ahead with the EU AI Act, which is the focus of most of the AI vendor community when it comes to understanding what AI regulation will likely shape model development.

The UK does not have standalone AI legislation or a new central AI regulator, and so companies must navigate AI compliance through existing laws. Cooper is arguing for a UN style approach to manage development and use of AI, saying:

On nuclear, international agreement came only after the world saw the terrifying power of the new technology at Hiroshima – and asked what would happen if it fell into the wrong hands. We cannot afford to wait for an AI equivalent of Hiroshima before we act.

Meanwhile the FCA is closely scrutinising its Mills review to assess how AI will re-shape financial services especially as it has found that a fifth of people across the UK would consider using AI to help them make financial decisions. This is concerning because financial regulators are not scrutinising AI models as part of their duties, which means that UK consumers have no right to compensation if they lose money as a result.

Added to these concerns are, of course, anxieties about loss of sovereignty because of reliance on foreign owned AI models from the US and China, which could lead to foreign censorship and data privacy vulnerabilities. Small wonder that regulated UK organizations are approaching the deployment of production-ready AI agents with caution. 

A sovereign-first AI offer for the UK

Atos is a very well-known and established service provider in the UK public sector and has already developed the Atos Sovereign Agentic Studio (SAS) that enables organizations to retain explicit control over where autonomy is applied, how decisions are governed and how data and models are operated across jurisdictions. According to the pitch, Atos SAS supports a production-first approach by enabling organizations to scale autonomy where it creates measurable value, while maintaining control where risk, regulation or criticality demand it.

Globally Atos SAS has Studios in the UK, France, Germany and the US, aligned with 10 global delivery centers. For the UK’s Sovereign Agentic AI capability, Atos will be using the Studio located in Birmingham in the UK where the agents will be deployed in mixed human and AI teams to enable sovereign-aligned AI for regulated operations.

Mike Hill, Head of Technology, Atos UK&I explains: 

 Our Sovereign Agentic Studio was built to turn AI ambition into operational reality and safely at scale. Kore.ai, as one of the leading providers of enterprise AI platform and solutions, is the ideal partner to extend that capability. Together, we are creating a unique, sovereign-first AI proposition for UK enterprises that demand both innovation and control.

Kore.ai brings its Agent Platform Artemis to the partnership. Typically, building complex AI agents that can “talk” to one another or integrate with third-party software takes months of custom programming. Artemis provides a way to compress this development cycle down to just days by using an AI engine to build, manage, and continuously optimize other AI agents. 

The salient characteristics of Kore.ai’s Artemis for this partnership include its coding language (Agent Blueprint Language) which enables constraints to be compiled directly into it, rather than relying on text prompts as guardrails. This is further strengthened by the platform’s deterministic runtime engine, a mechanism that intercepts actions in real-time. If an agent attempts to do something that does not conform to the specified rules, the engine stops it immediately. Additionally, the platform has a dual-brain architecture so that logical business rules are separated from the LLM and cannot be overridden or diluted by unpredictable agent activity. The platform is designed to create agents for governance and compliance, for highly controlled production level systems in regulated sectors.

As Raj Koneru, CEO and Founder of Kore.ai explains:

The next phase of enterprise AI will not be defined by individual agents, but by governed systems of intelligence operating reliably at scale. By combining the Kore.ai Agent Platform’s AI-native architecture, governance capabilities, and multi-agent orchestration with Atos’ Sovereign Agentic Studio, we enable UK organizations to build, scale, and optimize multi-agent AI systems within sovereign boundaries while maintaining the trust, security, and compliance required for regulated industries and mission-critical operations. This partnership helps enterprises move beyond AI pilots to production-grade AI with confidence.

My take

Will the partnership avert an AI equivalent of Hiroshima? Probably not, because we are relying on the morality of the people determining the rules. However, the combination of the Atos SAS production infrastructure, governance frameworks and UK industry expertise, with Kore.ai’s Artemis platform provides a compelling approach to the delivery of sovereign agentic AI for UK organizations.

Of course, it is by no means the only approach to combining deterministic business rules with the speed and efficiency of agentic AI. And many other sovereign AI offerings exist in the market for the UK. However, Kore.ai’s agentic platform is specifically designed for complex, multi-system workflows needing rigid governance, and other platforms for which this claim can also be made are often embedding AI directly into chat apps or deploying it straight into the underlying system of record. Kore.ai, on the other hand, relies on its multi-agent architecture and coding language ABL to weave disjointed systems together. Because of this it is a more flexible integration tool. 

The implementation effort is also greater than with other agentic platforms as business logic has to be scrupulously mapped. But this of course is where the partnership with Atos and its Atos SAS is crucial in offering UK organizations an option for moving the dial to production systems for agentic AI.