Analysis
The US and China in the AI race — which decisions remain with companies in Poland
The AI Business Center Foundation (AIBC) is a think tank and advisory partner that helps the boards of construction, industrial and technical-services companies adopt artificial intelligence so that it delivers measurable results and builds competitive advantage. This analysis concerns the rivalry between the US and China in AI and its consequences for the decisions of companies in Poland.
Choosing a model is no longer a technical decision. It has become a decision about dependencies
In July 2025 the United States announced a federal action plan on AI, “America’s AI Action Plan”, together with executive orders. The plan provided for removing regulatory barriers, accelerating investment in data centres, semiconductor plants and power grids, and promoting the export of the American AI technology stack: hardware, software and models. At the same time China was developing its own computing chips, state-backed industry alliances and the mass use of AI assistants in retail and customer service.
We describe decisions taken in mid-2025. The trade and export policy of both countries on AI technology changes frequently, so its current state must be checked before any purchasing decision.
What this means for a company in Poland
The rivalry between the major powers translates into three concrete decisions that remain with the board of a construction, industrial or technology company.
1. Which technology we tie ourselves to. Models and tools from different ecosystems differ in price, licence terms, where data is processed and the risk of trade restrictions. The choice of ecosystem affects the entire project portfolio, not a single tool.
2. Whether the company is ready to change provider. The risk of dependency does not arise from the choice itself but from the company’s processes being described in the language of a particular tool. A company with a process description that is not tied to one model can change provider when prices, terms or regulations change.
3. Which scenarios we include in planning. For technology purchases and development projects, it is worth considering scenarios of export restrictions, price changes and interruptions in access to cloud services.
The European framework remains
Regardless of the provider’s country of origin, AI systems placed on the market and used in the EU are in principle subject to the AI Act. For a company this means that the question of compliance with European rules is put to every provider, whether American, Chinese or European. In 2026 the AI Act was amended by Regulation (EU) 2026/1744, known as the AI Omnibus, and the transparency obligations of Art. 50 have applied since 2 August 2026.
The answer to market instability is order on the company’s side: semantic transformation, that is, organising a company’s names, processes and data so that people, language models and board reports understand them the same way. The model then becomes a replaceable component rather than the foundation.
What has changed since first publication
- We removed repetition and passages written for search engines that added no substance.
- We removed detailed descriptions of specific companies’ products whose current state we had not verified.
- We added a caveat on the volatility of export policy and a reference to the European legal framework after the AI Omnibus.
- We replaced the thesis that there are no “neutral” solutions with a practical principle: the model is replaceable, the process description stays.
Frequently asked questions
Can a company in Poland use Chinese language models?
This is a business and legal decision worth taking deliberately: where the data is processed, which licence terms apply, how the provider demonstrates compliance with the AI Act and which trade restrictions apply to the product at the time of purchase. The answer may differ for a cloud model and a model run locally.
How can a company limit the risk of depending on one ecosystem?
By describing its processes in a form not tied to a particular model, choosing the model for each task according to data, cost and security policy, and checking regularly that the model can be replaced without losing work.
Let’s start with a conversation
Let us talk about the dependencies your company already has in its AI tools.