Summary
- IBM surveyed 250 UK decision-makers at organisations with at least 500 employees.
- AI scored below the other sovereignty domains, and 41% reported an AI-related issue.
- Only 32% met IBM's strongest evidence standard, despite higher confidence in governance arrangements.
British organisations report strong confidence in their control over important digital systems, but many cannot demonstrate that their safeguards work under pressure, according to research published on 11 November by IBM.
The company’s Digital Sovereignty Index found that 41% of UK respondents had experienced an AI-related sovereignty issue during the previous three years. Only 32% met its highest evidence standard for demonstrating effective control, despite the UK’s relatively strong position among the seven markets surveyed.
The findings are based on 250 UK senior decision-makers at organisations with at least 500 employees, as part of a wider international sample of 2,003. They should not be treated as representative of all British businesses, and the assessment framework was developed by IBM rather than an independent regulator.
AI control scores trail other disciplines
IBM scored the UK at 77 out of 100 overall, compared with a seven-country average of 74. Its component scores were 81 for data, 80 for operations, 79 for technology and 70 for AI.
The company also identified a gap between confidence and verification. While 84% of UK respondents believed they could show accountability for AI systems, 27% reported having demonstrated that accountability through an audit, regulatory review, assurance activity or operational testing.
Such differences can arise when organisations establish policies without maintaining the technical records needed to show how systems operate. An AI application might have an assigned owner and approved purpose, but still lack a reliable account of which information it accessed or what actions it performed.
The problem becomes more demanding where AI agents interact with financial systems, customer records or external services. Permissions may be distributed across multiple applications, while responsibility for monitoring behaviour can remain unclear.
Digital sovereignty extends beyond hosting location
IBM defines sovereignty through the ability to retain and prove control across AI, technology, operations and data. This includes understanding responsibilities, tracing decisions, imposing constraints, intervening during incidents and recovering or leaving a service provider.
Where information is physically stored is one part of that assessment. A business can use a domestic data centre while remaining dependent on software, management systems or suppliers that it cannot readily replace.
In the UK sample, 93% of respondents reported at least one sovereignty challenge over three years. Among those affected, 83% had to provide formal evidence of their controls to an internal or external audience.
More than half of the affected group, 53%, said collecting evidence manually or through time-consuming processes made their response difficult. Fragmented records, untested exit arrangements and reliance on suppliers for information were also recurring obstacles.
These findings describe respondents’ reported experience rather than independently observed incidents. Nevertheless, they identify practical work that goes beyond changing a cloud contract or selecting a different hosting region.
Operational testing may reveal weaknesses
IBM found that 39% of UK respondents regularly tested their ability to demonstrate sovereignty, compared with 33% internationally. Of those who experienced challenges, 35% in the UK said they resolved their most demanding event without material disruption, compared with 30% across the sample.
The association does not prove that testing alone caused better outcomes. More mature organisations may also have stronger staffing, documentation and supplier management.
Testing can, however, reveal whether an organisation can recover data, move workloads or demonstrate what occurred when an automated system behaved unexpectedly. A policy asserting that a supplier can be replaced is different from having executed a successful migration.
Techopia recently examined the gap between cloud exit confidence and practical portability. IBM’s study broadens that concern to operational evidence across AI and other critical systems.
Evidence must survive organisational change
Systems and dependencies change regularly as businesses adopt services, add integrations and update AI applications. Controls verified during an annual audit may not reflect what software can access several months later.
Organisations seeking stronger evidence may need more accurate inventories, tested recovery procedures, records of automated decisions and ways to obtain relevant information from third parties.
IBM sells technology and consulting services in this area, so its commercial interest should be considered alongside the survey’s findings. The research does not establish that any particular IBM product solves the reported weaknesses.
The index offers a snapshot of how large organisations assess their control. The unanswered commercial question is whether businesses can improve verifiable accountability without creating an administrative burden that outweighs the resilience gained.












