Watchdog Polska Proposes Registry of AI Use in Public Offices

At a meeting of the Senate infrastructure committee on June 24, Michał Zemełka of the civic group Sieć Obywatelska Watchdog Polska (Citizens' Network Watchdog Poland) presented a proposal to create a public registry of artificial intelligence systems used by Poland's public administration. The proposal came just as the Senate was working on the law on artificial intelligence systems implementing the EU AI Act, giving the organization a rare chance to write its transparency demand directly into the legislation.

What Watchdog Is Proposing

The organization wants two measures written directly into the law. The first is a public registry where citizens could check which institution uses which AI system, in what area of activity, and for what purpose. The second is a requirement to flag the involvement of artificial intelligence in decisions, letters and other official documents whenever a model took part in preparing an analysis, recommendation or ruling.

Zemełka argued that transparency is not an obstacle to rolling out new technologies in public administration, but a precondition for trust in them. Watchdog has been repeating this demand since May, when it also raised it during consultations on the Policy for the Development of Artificial Intelligence in Poland until 2030 document, but only the work on the law gave it a real chance of being written into the regulations.

A citizen has the right to know whether technology affected how their case was handled - Michał Zemełka, Sieć Obywatelska Watchdog Polska

Wide Gaps Between Offices

The survey underpinning Watchdog's demand shows huge disparities in AI use between types of institutions. Large cities with county rights report using artificial intelligence in 34 percent of cases, central government offices in 20 percent, county administrations (starostwa) in 14 percent, and ordinary municipalities in under 1 percent. The organization also points to a problem with blurred definitions of AI, some institutions count only chatbots as artificial intelligence, leaving out document analysis systems, image recognition or decision-support tools.

Among the uses that offices themselves reported in response to information requests were transcribing recordings and summarizing meetings, working with documents and text analysis, chatbots and voicebots for residents, medical diagnostics, analysis of images and surveillance footage, and support for administrative decisions. Only 20 of the more than 2,000 municipalities surveyed have written internal rules for employees' use of AI, which Watchdog identifies as the main source of risk, not the use of the tools itself.

Law Nearing the Finish Line

The proposal came at a critical point in the legislative process. The government bill reached the Sejm in April, and the chamber passed the law on June 11 by a vote of 421 to 3, with 18 abstentions. The Senate submitted amendments on June 25, voting for them unanimously with 79 votes, after which the document returned to the Sejm, which completed its parliamentary work on it. The law implements the EU AI Act and establishes the Commission for the Development and Security of Artificial Intelligence, which will oversee the AI market in Poland and will be able to grant selected entities approval to test new technologies in regulatory sandboxes.

The government's reaction to Watchdog's proposal was positive, though so far only declarative. Deputy Digital Affairs Minister Dariusz Standerski said during the committee meeting that he supports the solution proposed by the organization and is ready to consult on the details with it and other parties. That does not mean, however, that the registry will make it into the final text of the law, at this stage of the legislative process, changing the scope of the provisions would require another round of amendments.

Why It Matters for Citizens

Without a registry and a labeling requirement, it is currently hard to establish whether a given administrative decision, for example on a benefit, a permit or an appeal, was made with the involvement of a language model or a decision-support system. Watchdog Polska has spent years working on the transparency of public institutions, including running the Sieć Sygnalistów (Whistleblowers' Network) and monitoring access to public information, so it treats AI as another area where offices should be accountable for the tools they use in dealing with citizens.

For Polish public offices, this would mean a new reporting obligation comparable to contract registries or registries of personal data sets. Institutions that have already deployed AI without internal procedures would have to build their documentation almost from scratch, given that only a handful of municipalities currently have written rules for employees' use of these tools.

The proposal's fate now depends on whether the issue resurfaces during work on the law's implementing regulations or in the activities of the newly created commission overseeing the AI market. Watchdog says it will keep watch on whether the deputy minister's declarations translate into a concrete legal provision, and plans further rounds of monitoring public offices in the coming years.