The growing sovereign artificial intelligence (AI) market is facing questions over transparency and accountability, with nearly a third of providers failing to disclose evidence supporting their sovereignty claims, according to a report from market intelligence firm Omdia.

The report, titled ‘Sovereign AI. A Playbook for Success’, assesses the contractual terms, service-level agreements, data processing agreements, privacy practices and certifications of 35 sovereign AI providers across 20 countries.

The providers examined include frontier AI laboratories, hyperscale cloud companies, data centre operators, neocloud providers and telecommunications operators.

Omdia found that sovereignty is frequently used as a marketing term to describe locally based infrastructure, without sufficient evidence that providers can guarantee the broader protections associated with sovereign services.

Almost a third, or 28 per cent, of the providers examined disclosed no evidence demonstrating their sovereignty credentials, while some expected enterprise customers to pay for audits to verify their claims.

The research also found that eight in ten contracts stated that network resilience fell outside providers’ control or responsibility.

Two thirds of providers limited compensation for service failures to service credits, while vague commitments to make “best efforts” were widespread.

The findings also challenged the assumption that local providers necessarily offer stronger sovereignty guarantees, with Omdia reporting that US-headquartered AI providers offered greater service transparency and better contractual terms than most local sovereign AI providers.

Meanwhile, providers in India were a notable exception to this trend, the report showed.

“No one is regulating who can call themselves a sovereign AI provider, and it shows,” said Camille Mendler, research director for telecommunications business-to-business services at Omdia.

“Local hosting is not sovereignty. Serious providers must demonstrate their credibility through transparent operations, fair contracts and proof of competence,” she continued.

“That includes the network, because data in transit must be sovereign too,” Mendler added.

The debate has implications for corporate technology spending, as sovereignty and regulatory compliance increasingly influence decisions about suppliers.

Up to 45 per cent of large enterprises plan to add or replace an information and communications technology (ICT) supplier within 18 months for sovereignty or compliance reasons, according to the research.

One in five large and multinational enterprises already requires network sovereignty safeguards, with these businesses ranking telecommunications operators as their most trusted AI partners.

Smaller and medium-sized enterprises could face greater risks because they often operate with limited IT teams and lack the resources needed to scrutinise providers’ sovereignty claims.

Omdia recommended that providers invest in certification and trust centres to demonstrate their credentials more clearly.

It also called on providers to publish separate service-level agreements for connectivity and AI services, allowing customers to distinguish between the commitments governing each component.

Providers should also seek partnerships to address gaps in their expertise and ensure they can deliver the protections associated with sovereign AI services.

The findings highlight the distinction between hosting data locally and providing demonstrable sovereignty guarantees across infrastructure, connectivity, contractual obligations and operational practices.