English Section

Poland yet to name AI watchdog as EU rules take effect: expert

29.08.2025 22:45
Poland has yet to designate the market surveillance authority required by the European Union’s Artificial Intelligence Act, a delay that could invite infringement proceedings, an expert has warned.
Pixabay License
Pixabay LicenseImage by Gerd Altmann from Pixabay

The law entered into force on August 1, 2024, with staged application dates, and EU countries were due to notify Brussels of their watchdogs by August 2, 2025, Polish state news agency PAP reported.

Legal expert Maria Dymitruk said that, despite the missing authority, Polish companies must already comply with the regulations where obligations apply, and violations can be penalised.

Dymitruk warned that the gap creates uncertainty, since some market players may wrongly assume the rules do not yet bind them.

Deputy Prime Minister and Digital Affairs Minister Krzysztof Gawkowski said he is not concerned about possible EU action, arguing Poland is helping set the pace on implementation and that many member states are still finalising their rules.

He said the government "will do everything to implement the Act quickly and well."

Under Poland’s draft Artificial Intelligence Systems Act, the Commission for the Development and Safety of Artificial Intelligence (KRiBSI), would serve as the market surveillance body.

Successive drafts were issued in October 2024 and February 2025, with the latest dated June 27, 2025, now before the European Affairs Committee.

Key deadlines have already passed. Bans on certain AI uses and a duty of “AI literacy,” meaning basic staff training and internal procedures for safe AI use, have applied since February 2, 2025.

Obligations for general-purpose AI models, known as GPAI and covering systems such as GPT-style large language models, took effect on August 2, 2025. Most rules for high-risk systems start in August 2026, and the Act is fully applicable in 2027.

GPAI providers must supply downstream developers with documentation that enables safe integration, publish summaries of training data, and adopt a copyright-compliance policy.

The European Commission and its AI Office supervise these obligations at EU level.

Penalties are significant. Breaches of banned practices can draw fines up to EUR 35 million or 7 percent of global turnover, whichever is higher.

Other serious violations, including failures tied to GPAI and high-risk systems, can lead to fines up to EUR 15 million or 3 percent of global turnover. Member states must also establish national sanction regimes.

(rt/gs)

Source: PAP