Where your data lives, and who decides what AI may do.
How we approach hosting, data sovereignty, security and AI governance. Our systems are designed around your infrastructure and operational requirements, with access controls, human oversight and accountability built into the architecture.
Your infrastructure. Your deployment requirements.
AI applications should fit your infrastructure, security requirements and data-processing obligations, not force you into a single hosting model.
- 01
Deployment on your terms
We support managed cloud, customer-controlled cloud and on-premises deployment, with the architecture selected around your existing infrastructure, integrations and operational requirements. In a managed deployment, Data Nucleus processes data on your instructions; in customer-controlled, on-premises and offline deployments, processing stays inside your own environment.
- 02
UK, EU or US hosting
Hosting can be arranged in UK, EU or US regions, according to customer requirements, applicable regulations and data sovereignty considerations. The selected cloud provider, services and region are agreed for each deployment. Managed deployments are hosted in the UK by default.
- 03
Data residency and processing
Hosting location alone does not determine where information is processed. We consider application hosting, data storage, backups, model inference, logs and external services when designing the deployment architecture.
- 04
Private and self-hosted AI
Where supported by the application and selected models, we can design deployments using private infrastructure or self-hosted AI models, reducing dependence on externally hosted inference services.
- 05
Integration with existing systems
Applications connect to the databases, document stores, ERP, CRM and operational systems already in use. Integration architecture is designed around the customer’s access requirements and existing technical environment.
- 06
Architecture matched to requirements
We select models, infrastructure and deployment patterns according to performance, cost, security, data-processing requirements and the capabilities of the selected providers.
- 07
Fully offline and air-gapped
Air-gapped deployment, with no outbound connection, where the application and selected models support it.
AI proposes. People and controls decide.
AI applications and agents should operate within defined permissions, technical boundaries and levels of human oversight. We design these controls around the workflow, the systems involved and the consequences of each action.
- 01
Authority defined by action
AI can recommend, prepare actions for human approval or automate defined tasks within agreed limits. The level of authority is determined by the workflow and the potential consequences of each action.
- 02
Controls beyond prompting
Role-based permissions, input validation, tool restrictions and downstream system controls limit what AI applications and agents can access and change. Security boundaries are enforced through technical controls, not prompts alone.
- 03
Human oversight where it matters
Workflows can route uncertain results, exceptions and consequential actions to authorised people for review. Approval points are designed around operational requirements and the level of risk involved.
- 04
Identity and access
Single sign-on and role-based access controls help ensure users can access the information and functions appropriate to their responsibilities. Permissions are configured around the organisation’s identity systems and access requirements.
- 05
Records and audit
Approvals, corrections and workflow actions can be logged to support traceability, operational review and investigation. Audit records are designed around the events and evidence relevant to each application.
- 06
Your choice of AI engine
We select and integrate AI models according to performance, cost, security, deployment options and data-processing requirements. Model selection is not tied to a single provider or architecture.
- 07
The control plane
The Data Nucleus control plane checks each proposed action against the authority approved in the agent’s Agent Risk Passport, and records each decision.
Security by design. Accountability in practice.
Our build approach draws on established security and data-protection principles, with controls selected according to the system’s intended use, deployment environment and applicable obligations.
- 01
UK GDPR and data protection
We design applications with data protection requirements in mind, including access control, data minimisation, retention and secure processing. Responsibilities and controls are defined according to the customer relationship and the personal data involved.
- 02
Security engineering principles
Our approach draws on NCSC guidance and established secure development practices. We consider authentication, authorisation, data flows, application interfaces and relevant threats during system design and implementation.
- 03
Governance and accountability
We define appropriate responsibilities, permissions, approval processes and audit requirements for each application. Where AI agents interact with external systems, their permitted actions and operating limits are part of the technical design.
- 04
Customer-specific assurance
Security and compliance requirements vary between organisations and sectors. We work with customers to identify relevant obligations, document the proposed technical controls and support the assurance activities required for their deployment.
Data Nucleus Ltd
Data Nucleus Ltd is a UK-registered company based in and operating from Manchester, England.
- Company registration
- 15379792
- Registered in
- England and Wales
- Website
- datanucleus.dev
Need the detail for a security review? We send the technical controls summary for the deployment model you are considering.
Request the technical controls summary