Artificial Intelligence in Gynecology and Obstetrics: Medico-Legal, Ethical, and Governance Challenges—A Narrative Review

Background: Artificial intelligence (AI) is entering gynecologic and obstetric imaging, risk prediction, decision support, robotic surgery, remote monitoring, and patient communication. An estimated 260,000 maternal deaths occurred worldwide in 2023, and in England obstetric claims account for 11% of clinical negligence claims but 55% of their value. AI errors, bias, or opacity in this specialty therefore carry major clinical and legal consequences. Objective: To synthesize the principal medico-legal and bioethical challenges associated with AI in gynecology and obstetrics and to propose a lifecycle governance framework for responsible clinical adoption. Methods: A narrative review with thematic synthesis was conducted. PubMed, Scopus, and Google Scholar were searched for records published from January 2020 to August 2026, supplemented by citation chaining and official legal databases. Two reviewers screened 131 records against predefined eligibility criteria; 41 sources (36 publications and 5 legal instruments) were included and analyzed thematically. Selection is reported in a PRISMA-style flow diagram. Results: Six connected risk domains were derived from five prespecified themes, namely validation and generalizability; professional competence and automation bias; distributed liability and evidentiary opacity; informed consent and patient autonomy; privacy and cybersecurity of reproductive data; and discrimination, access, and reproductive justice. Generative AI adds risks of hallucination, non-reproducibility, persuasive framing, and undocumented use. Existing law does not eliminate clinician responsibility: accountability is shared across developers, institutions, and professionals, but clinical judgment and documented oversight remain central. In the European Union, Medical Device Regulation and GDPR obligations already apply, whereas AI Act requirements for AI-enabled medical devices apply from August 2028 and the revised Product Liability Directive covers products placed on the market after 9 December 2026; causation and professional fault remain governed by national law. Conclusions: Safe adoption requires more than technical accuracy. We translate the findings into seven actionable institutional requirements, namely accountable ownership, minimum local validation, role-specific training, auditable documentation, equity monitoring, post-deployment surveillance, and incident reporting. AI should augment—not displace—clinically accountable and relational care.