Secure AI begins with trusted technology
Artificial intelligence dominated this year’s Labour Party Conference in Liverpool, with ministers, businesses and technology leaders united in wanting AI to boost productivity, improve public services and drive growth.
But amid the excitement about models, agents and automation, one vital question receives too little attention: Can we trust the devices and infrastructure AI relies on?
AI cyber security discussions often focus on models, data and governance, while overlooking the technology and infrastructure that enable AI to be deployed and run.
The Trojan Horse offers an enduring lesson: the threat was not forced inside; it was invited in.
Every device has a history of sourcing, manufacturing, transport, assembly and configuration. If that process cannot be verified, risk may enter before any AI application is deployed. As AI adoption grows, endpoints become more valuable targets - gateways to sensitive data, AI workloads, intellectual property and critical systems. Yet many organisations consider cyber security only after deployment.
Cyber risk now extends upstream across the technology supply chain, from sourcing and manufacturing to logistics and distribution. Trust must be verified, not assumed.
AI resilience requires AI at the edge
Device integrity also matters strategically for the UK’s AI future.
Although cloud infrastructure will remain essential, AI is becoming more distributed, running directly on PCs, workstations and other intelligent devices. AI at the edge can reduce latency, improve reliability, sustain operations when connectivity is disrupted, and give organisations greater control over sensitive data. For the UK, distributed AI also reduces dependence on a few centralised systems, creating a more resilient digital ecosystem, just as diversity strengthens energy, telecoms and other critical infrastructure.
A successful AI economy will combine world-class data centres with secure, intelligent endpoints that run workloads closer to users, widening access to innovation.
In distributed AI, security and resilience begin at the edge - not just in the cloud. AI security must therefore cover the entire technology stack, from models and AI applications to hardware, firmware and supply chains. The UK’s AI leadership depends on trust as well as innovation. Businesses and government will adopt AI only if the underlying technology is secure and resilient.
Organisations need not manage this alone. Trusted technology partners should demonstrate security from design and production through deployment, management and retirement.
When investing, organisations should ask: Can provenance be verified? Are manufacturing and chain-of-custody secure? Is hardware protected through delivery and refurbishment? Is there evidence, not just assurance? - These questions matter as AI becomes embedded in everyday operations.
Successful organisations will recognise that security starts before the first prompt, model or device reaches an employee. As the UK accelerates AI adoption, trust must underpin innovation. The organisations that can verify trust throughout the technology lifecycle will be best positioned to adopt AI securely.
