A blueprint for success: How Danes je nov dan’s advocacy led to a commitment for a Public AI Registry in Slovenia

In Slovenia, the use of artificial intelligence (AI) by the public sector has been expanding without significant public oversight, creating the potential for harmful or even dangerous uses of AI systems. To close this gap, EDRi affiliate Danes je nov dan launched an advocacy campaign calling for a national AI registry to ensure transparency and accountability. Their efforts led the Ministry of Digital Transformation to commit to establishing such a registry, a model that can inspire AI transparency initiatives across Europe.

By Danes je nov dan (guest author) · October 16, 2025

The growing need for AI transparency in the public sector

In recent years, the advancement of AI technologies led to their increased adoption within the Slovenian public sector. Despite numerous evidence that algorithms can produce biased results and discriminatory and dangerous outcomes, their use by the state has not received significant attention from the public or policymakers. This lack of oversight is a serious concern for fundamental rights.

Meaningful public debate on the automation of public services, substantial oversight over uses, and subsequent advocacy are only possible if the public has access to relevant information about which AI systems are in use. Yet, in Slovenia, it has been practically impossible to gain insight into the AI tools used by the state, how they function, or what is their purpose. Instead, fragmented efforts of individual CSOs and media have tried to fill the gap, often without the necessary resources to obtain all essential data.

How advocacy sparked government action

To address this lack of transparency, EDRi affiliate Danes je nov dan took the initiative and launched its own Public Sector AI Registry, offering the first centralised source of information on algorithms used by the Slovenian public sector. The registry currently lists nine systems, revealing how they are used, who developed them, how much they cost, and whether key safeguards – such as human rights or data-protection impact assessments – have been carried out. Despite the mapping, it is estimated that several dozen more are in use and are yet to be documented. Acknowledging that such efforts cannot replace systemic regulation, the organisation formally called on the Slovenian government to take action. It published a set of clear recommendations urging the Ministry of Digital Transformation and the Government of the Republic of Slovenia to establish an official, publicly accessible registry of all AI systems used in the public sector.

In a significant victory for transparency, the Ministry of Digital Transformation responded to this advocacy by publicly committing to establish such a registry, and has since included this commitment in the amendments to the national law implementing the EU AI Act.
This success demonstrates how targeted advocacy can lead to concrete government action, serving as a replicable model for organisations in other countries.

However, a public commitment is the first step. The organisation will continue to monitor the ministry’s activities and apply pressure until the promise is fulfilled and the registry becomes a fully operational tool for public oversight.

Key recommendations for an effective AI registry

For a registry to become an effective tool for democratic oversight, a public AI registry must be designed thoughtfully and inclusively. Danes je nov dan proposed that the state pursue the following objectives in its design:

  1. The public must have access to comprehensive information about AI systems.
    Transparency cannot end with merely publishing a system’s name. For effective oversight, a registry must include crucial data about the algorithm’s entire life cycle, from procurement information and financing sources to a clear explanation of its intended use and its role in decision-making. It is also essential to include technical details about the model’s developer, the datasets used for training, and information on available complaint mechanisms.
  2. The public must be informed about all AI systems in the public sector.
    The registry must include all AI systems in use, not just those defined as “high-risk” under the EU’s AI Act. The EU database for high-risk systems will exclude many systems from public view, such as those used for law enforcement as well as in the areas of migration and border control. A national registry should also include systems that are still in development or already decommissioned, providing a complete picture.
  3. The registry must be user-friendly and accessible.
    Information is only useful if it is accessible and understandable. The registry must be designed to allow for easy searching and filtering, and all changes to its entries must be logged. To enable further analysis by experts and the media, the database should be machine-readable and accessible via an API, and the registry’s source code should be open.

A replicable model for democratic oversight

Danes je nov dan adds that to realise this vision, the Slovenian government must establish a clear legal framework, ensure stable funding, and appoint responsible individuals for maintaining the registry. Crucially, the development and oversight process must involve external stakeholders, especially civil society organisations that represent the rights of vulnerable groups. The Slovenian case provides a valuable blueprint for how strategic advocacy can secure government commitments, creating a foundation for greater democratic accountability and the responsible use of new technologies.

Contribution by: EDRi affiliate, Danes je nov dan