A Drop-In Replacement, Not Just A Pitch
Denmark's DCAI (Danish Center for AI Innovation) and Corti, a Copenhagen-based clinical-AI company, announced Corti Models on August 20, 2026: a governed AI infrastructure layer giving European enterprises sovereign access to leading AI systems through a single, OpenAI-compatible API, deployable on sovereign cloud or on-premises. "European enterprises need access to world-class AI without depending on a single provider," said Corti CEO Andreas Cleve.
The compatibility detail is the whole point. Most sovereignty pitches this year have been rhetorical: fund a domestic lab, pass a disclosure law, announce a cloud region. This one is mechanical. An enterprise's existing code, already written to call OpenAI's API format, needs no rewrite to route through sovereign infrastructure instead, which is the actual barrier that usually stalls a switch, not the availability of an alternative.
Built On Denmark's National AI Infrastructure
The layer runs on DCAI's Gefion supercomputer, ISO-certified, NVIDIA-powered infrastructure operating under European governance and security standards, which DCAI is now expanding with next-generation NVIDIA Blackwell B300 clusters for large-scale enterprise and public-sector AI workloads across the Nordic region. "Sovereign AI infrastructure transforms into practical enterprise solutions creating real value," said DCAI CEO Martin Svensson.
Corti itself is not a generalist newcomer: its flagship clinical model, Symphony, already serves more than 100 million patients a year across health systems, giving Corti Models a healthcare-grade governance baseline before it takes on a broader enterprise client base. Trifork, a 1,102-employee software firm operating in 16 countries, is the first implementation partner, tasked with deploying Corti Models for enterprise clients while holding governance and compliance standards. "Customers want digital independence without compromising security, compliance, or data control," said Trifork CEO Jorn Larsen. DCAI and Corti cite a potential 5-to-10x reduction in AI spend from centralizing model access, data handling, and billing under one contractual relationship instead of per-team vendor sprawl.
Part Of A Pattern, Not A One-Off
This launch lands the same week Servola reported France excluding OpenAI from state security-testing tools in favor of Mistral, and Cisco naming sovereign clouds explicitly as a target customer for its expanded AI infrastructure line. Three separate announcements, three different countries, all answering the same question EU governments have spent 2026 asking out loud: how does an enterprise get frontier AI capability without a single foreign vendor holding the data, the pricing, and the access decision.
The Part That Actually Matters For Regulated Sectors
Most sovereign-AI announcements this year have been supply-side signals with no clear adoption mechanism attached, funding for a lab, a cloud region opening, a policy paper. Corti Models is different because the adoption cost is close to zero: if your integration already speaks OpenAI's API dialect, which most enterprise AI tooling built in the last two years does, pointing it at sovereign infrastructure instead is a configuration change, not a migration project.
For regulated EU sectors carrying GDPR, NIS2, or DORA exposure on foreign AI vendor dependence, that compatibility layer is the difference between a sovereignty option that gets evaluated in a slide deck and one that gets tested in a procurement pilot. The claim of a 5-to-10x AI spend reduction from centralized governance is Corti and DCAI's own figure, unverified by a third party, and worth treating as a vendor claim rather than a benchmark until an independent enterprise deployment reports real numbers.
Read next: Anthropic Lets You Host Its Mandatory Logs Yourself | Sovereign Cloud Does Not Mean What You Think It Means
