(IN)VALIDADE DO RECONHECIMENTO FACIAL COMO PROVA: RISCOS, VIESES E PARÂMETROS

Resumo

The incorporation of facial recognition systems in criminal prosecution has intensified in recent years, driven by advances in artificial intelligence and the growing use of biometric technologies by public security agencies. Although presented as instruments capable of increasing investigative efficiency, these systems raise controversies regarding their technical reliability, neutrality, and compatibility with the structuring principles of criminal procedure, especially in light of evidence of errors, biases, and regulatory shortcomings. Therefore, this study aims to investigate the (in)validity of facial recognition as a means of proof in criminal proceedings. Methodologically, this is a qualitative and exploratory study, developed through a systematic review of the literature published between 2020 and 2025. The searches were conducted in the Scielo, Capes Journals, and Google Scholar databases, resulting in the final selection of seven empirical and theoretical-applied studies, after screening by title, abstract, and full reading. The findings indicate that facial recognition performance is highly dependent on usage conditions and exhibits significant error rates in real-world forensic contexts, as well as systematic biases that affect specific groups. It is concluded that facial recognition does not constitute conclusive proof and should be admitted, at most, as an auxiliary element, subject to rigorous legal parameters and safeguards compatible with due process.