With AI being integrated into the UK’s public services and national infrastructure at almost every level, the question of who controls the underlying technology has never been more urgent.
Among the various issues at play for the government to grapple with are rising infrastructure costs, vendor lock-in and concerns about digital sovereignty. AWS and Microsoft, both US-owned, currently control up to 80% of the UK cloud services market, and it seems likely that the debate over where the control really lies will continue for some time to come.
In fact, the current direction of travel looks set to further embed US tech giants within domestic AI infrastructure. As a backdrop to President Trump’s recent State Visit, the two governments announced £150bn of US tech investment, a significant proportion of which is geared towards AI.
Behind the headlines, however, there are some difficult challenges to address. How, for example, can public sector organisations gain access to the kind of AI infrastructure that will deliver the transparency required in real-world environments? Also, how can they implement the appropriate levels of internal governance and privacy to ensure AI can meet the growing range of regulations coming into effect
Equally important is the question of impact effectiveness. Research indicates that, in a business context, the share of companies abandoning most of their AI initiatives has jumped to 42%, up from 17% last year. Looking elsewhere, the figures are even more dramatic, with a study from MIT finding that 95% of generative AI pilots are failing because, as one commentator put it, they are “. . . slick enough for demos, brittle in workflows”, with projects “stuck in high-adoption, low-transformation mode.” For public sector organisations, where failed IT projects have become a perennial and embarrassing occurrence, public support for AI investment will depend on delivering value.
The case for open source AI
So, what’s the answer? Part of the solution lies in rethinking the infrastructure AI is built on. If the current generation of closed systems is making it harder to meet regulatory expectations, then the case for open models that prioritise visibility, auditability and ownership requires closer consideration.
On a fundamental level, open source principles, which are based on issues such as collaborative improvements, knowledge pooling, and transparency, are well-suited to creating an AI ecosystem that balances innovation with ethics. Many would argue that this is essential if governments are to avoid problems associated with vendor lock-in, limited transparency and restricted access to foundational models associated with uncompetitive technology markets.
If, however, AI development remains the domain of relatively few dominant players, users will undoubtedly face higher costs and will be offered fewer opportunities to tailor solutions to their specific needs. This situation will only intensify as more AI is integrated into public sector infrastructure.
In contrast, enterprise-ready open-source AI solutions are built to offer organisations the control, flexibility and security required to deploy and manage workloads at scale. This includes the power to integrate preferred large language models (LLMs) while maintaining full ownership of the associated data and infrastructure. The range of open-source AI alternatives is growing rapidly, with DeepSeek being perhaps the most well-known open-source AI technology. It drew worldwide attention to the potential of underlying AI models that are freely available for use, modification and distribution.
At the scale required for public sector use cases, technologies are being developed that can be securely deployed across cloud, on-premises, hybrid, or air-gapped environments. With predictable costs and extensible architecture, open-source AI platforms help organisations avoid vendor lock-in while adapting to evolving needs. At the same time, they also support compliance and observability – vital components to ensure control remains in the hands of each organisation where the technology is being implemented.
The momentum behind open source AI is growing. A 2025 McKinsey survey found that over 50 per cent of respondents were using an open source solution in each of the data, models, and tools areas of the technology stack. In addition, more than three-quarters of respondents expected their organisations to increase the use of open source AI technologies during the next several years. As the report summary concludes, “Much like in the cloud and software industries, a multimodel approach will likely be prevalent for many companies, with open source and proprietary technologies coexisting in multiple areas of the AI technology stack.”
Mark Dando
Mark Dando is General Manager for EMEA North at SUSE. An enterprise software leader with a strong track record in growth and transformation, Mark has led major changes, including a GTM overhaul and a regional relaunch, helping Europe’s leading businesses modernise with SUSE’s cloud-native tools.
