President Karol Nawrocki signed the underlying law on Friday.
It takes effect in about two weeks and will let firms trial AI tools in a controlled environment before rolling them out.
Deputy Digital Affairs Minister Dariusz Standerski said the sandbox, described as the first of its kind in the EU, should be running within three to four months.
Companies will be able to check if their AI systems meet the rules without facing penalties during that early testing period, Standerski explained.
Buyers, meanwhile, "will get assurance that AI products on the market have already been tested and found safe."
Digital Affairs Minister Krzysztof Gawkowski, who also serves as the country's deputy PM, said the new law marked an important step for AI in Poland, helping both to support new technology and improve oversight of it.
The legislation also creates a new watchdog, the Commission for the Development and Safety of Artificial Intelligence, to oversee AI regulation in the country.
It will run the sandbox, investigate problems with AI systems and vet high-risk tools used in areas such as education, hiring and critical infrastructure.
The body will have the power to fine companies or pull non-compliant systems from the market.
The commission is to be chaired by an official appointed by parliament for a five-year term, due to be in post by October, with the rest of the commission following in November.
The EU’s AI Act took effect in August 2024, with rules being phased in gradually.
A ban on certain AI systems came into force in February 2025, national watchdogs were required by last August, and a requirement to label AI-generated content begins this August.
The law applies in full from 2028.
(ał)
Source: PAP