Last updated
In healthcare, the ability of artificial intelligence to analyse large volumes of data quickly and accurately, to identify complex patterns and to read images offers real possibilities: personalised treatment plans, greater diagnostic accuracy, more patient-centred care. In reproductive medicine it is a powerful tool that improves the chances of a pregnancy and optimises fertility treatment. The main directions in which algorithms are being used are the assessment and selection of reproductive cells; embryo selection; personalising ovarian stimulation; predicting success; personalised treatment plans and better diagnostic accuracy; improving workflows; producing standardised protocols, patient information forms and standard operating procedures; and monitoring fetal development for possible abnormalities.
Ukraine is no exception to the trend, and Ukrainian fertility clinics have been bringing these algorithms into practice. The IVMED centre, for example, uses two systems, ERICA™ and SiD from IVF 2.0, which process volumes of data a person physically cannot and remove the subjective human factor, helping to choose the best sperm for fertilisation and the most promising embryos for transfer.
Against that optimism stands a simple fact: artificial intelligence is a relatively new term for the Ukrainian legislator, and current Ukrainian law contains neither a settled legal definition nor, still less, substantive regulation. That is the essential starting point — no field, not just reproductive medicine, has any meaningful body of AI regulation to build on, so this has to begin almost from nothing.
A regulatory base is nonetheless taking shape. The Law on personal data protection addresses the question in part. In October 2019 Ukraine, a member of the Council of Europe’s Ad Hoc Committee on Artificial Intelligence, joined the OECD Recommendation of the Council on Artificial Intelligence. In 2020 the government approved the Concept for the development of artificial intelligence (Cabinet Order No. 1556-r of 2 December 2020), which for the first time in Ukraine defines the term: an organised set of information technologies with which complex tasks can be performed through scientific research methods and algorithms for processing information obtained or independently created in the course of work, and which can create and use its own knowledge bases, decision-making models and algorithms, and determine ways of achieving the tasks set. Among the problems the Concept identifies are the absence or inadequacy of legal regulation, low investment, insufficient quality of higher education in the field, the difficulty of checking that AI systems comply with the law and with ethical principles, and the lack of common criteria for what counts as ethical across different sectors.
Two further steps followed: the roadmap for regulating AI presented by the Ministry of Digital Transformation on 7 October 2023, and the White Paper published on 26 June 2024, which sets out Ukraine’s approach in stages. The first, preparatory stage is about non-legislative instruments to help business and other stakeholders get ready for future mandatory regulation — a regulatory sandbox, a methodology for assessing the impact of AI on human rights, labelling tools, and soft-law instruments such as voluntary codes of conduct and general and sectoral recommendations. Second, in the period before a binding special law exists, what matters is compliance with the rules that already apply, above all data-protection law — for instance the data subject’s right to protection against an automated decision with legal consequences for them. The absence of an AI law is no excuse for not complying with the law that does exist. The third stage is harmonisation with EU law. The EU AI Act, adopted on 13 March 2024 and in force since 1 August 2024, takes a risk-based approach across four categories and expressly classes AI systems used in healthcare as high risk, with the requirements that follow: registration, algorithmic transparency, mandatory human oversight, logging of AI decisions and auditability. For Ukraine that is the reference point.
The core of the article is the set of legal challenges at the meeting point of AI and assisted reproduction.
Leaks of personal medical data. The genetic information processed during treatment is the most sensitive category of personal data there is, and the Law on personal data protection contains no special provisions on processing it by algorithm. The Concept itself notes the inadequate level of information security in state systems, with outdated detection tools and unused capacity for anticipating threats.
The black box. Many deep-learning algorithms are opaque. In practice doctors sometimes cannot explain how the system arrived at its choice of a particular cell or embryo, which makes it hard to give patients reasoned explanations and creates serious barriers to trust, oversight and patient safety.
Algorithmic bias. Much of the research is retrospective, with few prospective randomised trials, which raises the risk of bias, departure from reporting standards, and missing data and code. And where an algorithm has been trained largely on data from one group of patients, it may be less accurate or produce irrelevant predictions for other ethnic groups, age categories or atypical cases. With no regulation of that bias, the risk of discrimination in the exercise of reproductive rights is real.
Informed consent. In Australia, women undergoing IVF with the Ivy algorithm were given full information about it and only then consented to its use. Ukrainian law has no equivalent mechanism of “digital informed consent”. Such a consent should tell the patient which AI systems will be used, their nature, purpose and expected duration of use; state that the data collected and processed will be protected under data-protection law; and include the doctor’s oral explanation of the risks and of alternative methods of treatment.
Liability. When an algorithm selects an embryo that proves non-viable, or misclassifies genetic material, who answers for it — the clinic, the software developer, or the embryologist who accepted the recommendation? Neither scholarship nor legislation gives a clear account of AI’s legal personality, which leaves medicine among the riskiest fields. Most countries still leave responsibility with the medical professional even where the diagnosis or recommendation came from a system, on the principle that the doctor has the last word and must assess the data critically. That creates legal uncertainty, since with complex diagnostic systems a doctor cannot in fact verify every conclusion; it does not encourage innovation, and it can shift responsibility unfairly. It seems logical that developers and manufacturers should answer where the error was caused by a technical defect, a flaw in the algorithm, bias in the training data or a breach of safety requirements. Ukrainian court practice takes the same line that AI is an auxiliary tool only: in a ruling of 8 July 2025 the Commercial Court of Cassation held that answers given by artificial intelligence are not recognised as a source of reliable, scientifically proven information, so refusing to examine them as evidence was lawful.
Ethics. Using AI to assess embryo viability could lead towards “designer babies”, with parents selecting not only for health but for desired characteristics such as intelligence or appearance — which stigmatises people with disabilities and raises substantial ethical objections. There is also the risk of delegating decisions of this weight to machines, reducing the doctor’s role as a person able to give emotional support and turning patients into a set of biological data.
What this analysis reveals is a deep structural split: the technology is already working and the law has not caught up. That is true of many fields, but it is especially acute here, because these decisions concern the most fundamental thing of all — the coming into being of a human life. The priority should be two connected laws: a law on assisted reproduction with a separate chapter on digital informed consent, liability for algorithmic error, standards for protecting genetic data and mandatory human oversight of AI decisions; and a law on artificial intelligence implementing the logic of the EU AI Act. In the terms of the White Paper, the mark on the scale between human rights and innovation should be set where patients are protected from the risks these technologies carry and from their malicious use.