How to Develop AI Products — Recommendations for Companies and Startups

 

The IT Ukraine Association has acted as an expert in the development of recommendations by the Ministry of Digital Transformation regarding the responsible development of technological systems with artificial intelligence. These guidelines will help companies and startups implement AI in their products and services without violating human rights, copyright or private property.

 

Demand for responsible AI is growing in the global market. Users of innovative products expect breakthrough AI solutions from companies that are safe for them. These recommendations serve as a practical guide for Ukrainian developers on how to maintain innovation in their products while designing solutions that respect human rights,

— said Oleksandr Bornyakov, Deputy Minister of Digital Transformation for IT Development.

 

The recommendations are based on the best global standards, including the HUDERIA methodology from the Council of Europe, which helps assess the impact of AI systems on human rights.

 

The document’s recommendations can be applied to the development of technological solutions across various sectors — finance, healthcare, education, employment, marketing and more.

 

It contains practical instructions for companies on selecting a large language model and adapting it through training and fine-tuning. Specifically, it is recommended to consider not only the number of parameters or the model’s popularity but also the quality of training data, transparency, adaptability and vendor reputation.

 

The document provides guidelines regarding the data used for training the model, including how to process it before training to ensure the model operates without hallucinations, discrimination or bias.

 

Additionally, experts compiled a list of key human rights and examples of violations in AI products. These cover cases in medicine, employment, education, finance and more. These examples will help companies avoid common mistakes during product development.

 

A separate section offers advice on avoiding copyright and property rights infringements, especially when using generative AI.

 

The appendix includes over 30 tools for AI solution development, detecting code errors, auditing bias and fairness in systems and more.

 

Key recommendations include:

 

• Use only datasets for which you have legal rights for model training, and ensure data are balanced, complete, anonymised and normalised.
• Involve users in product creation and improvement. Adopt an iterative approach where you regularly test the model, gather feedback and adapt the product. This helps make the system fairer, more useful, and accurate — especially in sensitive areas.
• Ensure expert human oversight, particularly in decision-making about access to services, loan approvals, employment, education and healthcare.
• Respect copyright. Always check the terms of use of models and services, and verify that generated content does not contain third-party copyrighted materials.
• Implement privacy by default and by design: minimise data collection, ensure encryption, access control, logging and the right to data deletion. Users must control their information.

 

---

 

The recommendations were developed by the Ministry of Digital Transformation together with the Expert Advisory Committee on AI Development at the Ministry, the IT Ukraine Association, the Ukrainian Bar Association, CEDEM,  the Department of Intellectual Property and Information Law  of the Educational and Scientific Institute of Law at Taras Shevchenko National University of Kyiv, and the Research Institute of Intellectual Property of the National Academy of Legal Sciences of Ukraine.

The document was developed with the support of the “Digitalisation for Growth, Integrity and Transparency” project (UK DIGIT), implemented by the Eurasia Foundation and funded by UK Development.