A Baltic data agency is buying and adapting AI models to upgrade institutional infrastructure, reflecting a wider shift towards AI-enabled public services.
Follow Tenderlake on LinkedIn for concise insights on public-sector tenders and emerging procurement signals.
Buyer VĮ Registrų centras (PV) has launched the Artificial Intelligence Solutions Procurement, seeking to acquire and adapt artificial intelligence models for institutional infrastructures. The move points to a shift from isolated pilots towards embedding AI in the systems that sit behind public services.
Published on 24th December 2025, the contract notice is unusually brief. It simply states that the acquisition and adaptation of artificial intelligence models for institutional infrastructures are being sought. There is no detail yet on particular departments, datasets or functions.
Even so, the language matters. The focus on 'models' rather than individual software tools, and on their 'adaptation' to existing institutional infrastructures, suggests more than a narrow use case. It points towards tailored AI that must fit into established technical environments, legal frameworks and ways of working across one or more public bodies.
The emphasis on adaptation also implies work beyond licensing technology. It indicates a need to shape generic AI capabilities so they can operate inside public-sector constraints, whether around data protection, transparency, auditability or security. Those themes are becoming common across government AI procurements in Europe.
The notice from VĮ Registrų centras (PV) lands in a year marked by a series of public tenders for AI platforms and environments, not just standalone applications. Across multiple jurisdictions, agencies are buying the underlying capability to develop and run AI, rather than one-off tools.
In August 2025, Ministerstwo Rodziny, Pracy i Polityki Społecznej set out a similar ambition in its AI Platform Implementation notice. That contract covers the delivery, installation, configuration and commissioning of a complete AI platform, including a server and software for operating artificial intelligence models. It is a deliberate attempt to create a standard environment for AI inside a ministry, rather than tackling each problem separately.
A similar pattern appears in financial oversight. On 21st July 2025, Urząd Komisji Nadzoru Finansowego signalled plans to expand its AI capability through the AI Platform Expansion prior information notice. That order covers delivery of a server and software licences, implementation of an expanded AI platform with security integration, consulting services, components for AI solutions, and the construction of two AI models.
On 20th October 2025, Narodowy Bank Polski went further, issuing an AI Environment Purchase to acquire an artificial intelligence environment that includes servers, a management platform, software, technical support and implementation services. This kind of self-contained 'AI environment' has become a template for institutions that want to control how models are developed, trained and deployed on sensitive data.
There is also a turn towards shared platforms. In November 2025, Staatskanzlei launched its AI Platform for Municipalities project, aiming to procure a platform for Zurich municipalities to enhance administrative efficiency, support workflows and give smaller municipalities easier access to AI technology. Crucially, the platform will centralise legal and IT security clarifications, relieving individual municipalities from having to solve those issues alone.
In this context, the procurement by VĮ Registrų centras (PV) looks like part of a broader effort to build AI into the fabric of public administration. Rather than focusing on a single problem, it frames AI as something that must be integrated into 'institutional infrastructures' as a whole.
Alongside platform and infrastructure buys, 2025 has seen a steady stream of tenders for concrete, often quite mundane AI applications. These are the kinds of tasks that could eventually sit on top of the models and infrastructures now being procured.
In November 2025, Valtion talous- ja henkilöstöhallinnon palvelukeskus sought an AI-Based Invoice Classification Solution capable of generating classification proposals for electronic invoices. The system is expected to learn continuously from processed data while meeting public administration requirements. It is a classic back-office use of AI: automating repetitive, rules-based work while keeping humans in the loop.
On 31st October 2025, Valstybinė ligonių kasa prie Sveikatos apsaugos ministerijos issued a contract notice for an Intelligent Chatbot Implementation, covering creation, installation, maintenance and support. Here, AI is being used to handle routine interactions, freeing staff time and offering round-the-clock access to information.
The focus on decision support is also growing. In September 2025, Úrad pre verejné obstarávanie launched its AI Utilization in Procurement Office project, which aims to implement AI technology to enhance decision-making efficiency and consistency within a public procurement office. And in November 2025, Poliisihallitus (National Police Board of Finland) sought IT expert services for AI-focused application development through its AI Application Development Services notice, covering work from definition to maintenance of police information systems.
Seen together, these contracts show AI moving into routine administrative and regulatory processes. The notice from VĮ Registrų centras (PV) sits a layer below that – focused on the models and infrastructures that could underpin many such applications over time.
Some organisations are still at the stage of buying hardware for AI workloads. July 2025 saw VRHOVNO SODIŠČE REPUBLIKE SLOVENIJE order a single AI Server System Purchase, and in September 2025 Universität Oldenburg followed with an AI Server tender for delivery, installation and commissioning of a GPU server for AI applications. Universities such as Semmelweis Egyetem and Universidade de Évora have similarly sought AI servers to support research and supercomputing.
The VĮ Registrų centras (PV) procurement sits a step higher in the stack. It is not framed as a hardware purchase, nor as a single, narrow application. Instead, it calls directly for AI models and their adaptation to 'institutional infrastructures'. In practical terms, adaptation typically involves training or fine-tuning models on institutional data, aligning them with internal policies, and integrating them with existing systems and workflows. It can also mean constraining models so they operate transparently and predictably in environments where accountability is critical.
That kind of work is resource-intensive and often iterative. It requires not just technical capability but close cooperation between technology teams, domain experts and those responsible for governance and oversight. The sparse wording of the notice leaves open whether this will be delivered as a one-off project, a longer-term partnership, or a platform that others in the public sector can later reuse.
For now, many key details of the Artificial Intelligence Solutions Procurement remain unspecified in the public text: the precise scope of the institutional infrastructures involved, the types of models envisaged, and how success will be measured. Future documentation and contract awards will show whether this becomes a central AI capability that others can build on, or a more focused institutional toolset.
What is clear is that it aligns with a broader shift in 2025: public bodies are no longer only experimenting with AI at the edges, but are beginning to procure the models, platforms and environments that sit at the heart of their operations. As similar projects in ministries, regulators, municipalities and agencies move from procurement to implementation, they will offer early clues on how effectively AI can be woven into public-sector infrastructure – and on the governance models needed to keep it trustworthy.
Follow Tenderlake on LinkedIn for concise insights on public-sector tenders and emerging procurement signals.