DEEPFAKES E PROVA DIGITAL NA VIOLÊNCIA DOMÉSTICA: OS RISCOS DA INTELIGÊNCIA ARTIFICIAL
DOI:
https://doi.org/10.14210/nej.v30n3.p398-419Palavras-chave:
Prova audiovisual, Valoração, deepfakes, autenticidade, Inteligência ArtificialResumo
Contextualização: Os avanços recentes da inteligência artificial generativa têm introduzido novos desafios ao sistema de justiça penal, em particular no domínio da prova digital. Entre as aplicações mais problemáticas destacam-se os deepfakes, suscetíveis de induzir em erro quanto à sua autenticidade, especialmente em contextos de maior fragilidade probatória, como os crimes de violência doméstica.
Objetivos: O presente estudo tem como objetivo geral compreender as perceções, experiências e práticas dos/as magistrados/as relativamente à admissibilidade e valoração da prova digital, com especial enfoque nos meios de prova audiovisual e nos riscos associados à sua eventual manipulação por tecnologias de inteligência artificial generativa.
Método: Foi adotada uma abordagem qualitativa, recorrendo-se à realização de entrevistas semiestruturadas a oito magistrados/as, com experiência profissional em processos de violência doméstica.
Resultados: Observaram-se limitações na deteção de elementos audiovisuais gerados ou manipulados por inteligência artificial. A autenticidade deste tipo de elementos (deepfakes) granjearam mais credibilidade que elementos autênticos. De todo o modo, embora a prova audiovisual seja considerada relevante, o depoimento da vítima continua sendo considerado a “prova rainha”.
Downloads
Referências
BELCIC, Ivan; STRYKER, Cole. What is GPT (generative pretrained transformer)?. IBM, [S.l.], [S.d.]. Disponível em: https://www.ibm.com/think/topics/gpt. Acesso em: 03 dez. 2025.
BRAUN, Virginia; CLARKE, Victoria. Thematic analysis. In: COOPER, Harris et al. (Org.). APA handbook of research methods in psychology. v. 2. Washington, DC: American Psychological Association, 2012. p. 57–71.
BRAUN, Virginia; CLARKE, Victoria. Using thematic analysis in psychology. Qualitative Research in Psychology, London, v. 3, n. 2, p. 77–101, 2006.
BRINKMANN, Steinar. The interview. In: DENZIN, Norman K.; LINCOLN, Yvonna S. (Org.). The SAGE handbook of qualitative research. 5. ed. Thousand Oaks: SAGE Publications, 2018. p. 997-1038.
CHESNEY, Robert; CITRON, Danielle K. Deep Fakes: A looming challenge for privacy, democracy, and national security. California Law Review, v. 107, n. 6, p. 1771-1786, 2019.
CHINA, Chrystal R. Five machine learning types to know. In: Cole STRYKER et al. (eds.). The 2025 Guide to Machine Learning. [S.l.]: IBM, 2025. Disponível em: https://www.ibm.com/think/topics/machine-learning-types. Acesso em: 01 nov. 2025.
COLLINS DICTIONARY. Collins Online Dictionary: definitions, thesaurus and translations. Disponível em: https://www.collinsdictionary.com. Acesso em: 03 dez. 2025.
COMISSÃO EUROPEIA. Comunicação da Comissão ao Parlamento Europeu, ao Conselho, ao Comité Económico e Social Europeu e ao Comité das Regiões: Roteiro para o acesso lícito e efetivo aos dados para efeitos da aplicação da lei. Bruxelas, 2025.
DIXON JR., Herbert B. DeepFakes: More frightening than photoshop on steroids. Judges’ Journal, v. 58, n. 3, p. 35-37, 2019.
DURÃES, Dalila; FREITAS, Pedro Miguel; NOVAIS, Paulo. The Relevance of Deepfakes in the Administration of Criminal Justice. In: ANTUNES, Henrique Sousa et al. (eds). Multidisciplinary Perspectives on Artificial Intelligence and the Law. Cham: Springer, 2024.
EDMONDS, Lauren. 3 years old, 800 million users, 29,000 prompts a second: ChatGPT's meteoric rise, by the numbers”. Business Insider, [S.l.], 2025. Disponível em: https://www.businessinsider.com/chatgpt-by-the-numbers-2025-11#chat-gpt-has-800-million-weekly-active-users-2. Acesso em: 03 dez. 2025.
EUROPOL. Facing reality? Law enforcement and the challenge of deepfakes. Luxemburgo: Publications Office of the European Union, 2022.
GAGLIARDI, Marília Papaléo. Artificial Intelligence and Women’s Rights. In: QUINTAVALLA, Alberto e TEMPERMAN, Jeroen (eds.). Artificial Intelligence and Human Rights. Oxford: OUP Oxford, 2023.
GOOGLE. What is machine learning (ML)?. Google, [S.l.], [S.d.]. Disponível em: https://cloud.google.com/learn/what-is-machine-learning?hl=en. Acesso em: 03 dez. 2025.
GOOGLE. What is Machine Learning? In: GOOGLE. Introduction to Machine Learning. [S.l.], 2025. Disponível em: https://developers.google.com/machine-learning/intro-to-ml/what-is-ml. Acesso em: 27 nov. 2025.
LEE, Fangfang. What is a neural network? In: Cole STRYKER et al. (eds.). The 2025 Guide to Machine Learning. [S.l.]: IBM, 2025. Disponível em: https://www.ibm.com/think/topics/neural-networks. Acesso em: 30 nov. 2025.
LEVY, Neil. Bad Beliefs: Why They Happen to Good People. Oxford: Oxford University Press, 2022.
MARAS, Marie-Helen; ALEXANDROU, Alex. Determining authenticity of video evidence in the age of artificial intelligence and in the wake of Deepfake vídeos. The International Journal of Evidence & Proof, v. 23, n. 3, p. 255-262, 2018
OPENAI. Fine-tuning GPT-2 from human preferences. OpenAI, [S.l.], 2019. Disponível em: https://openai.com/index/fine-tuning-gpt-2/. Acesso em: 03 dez. 2025.
OPENAI. Introducing ChatGPT. OpenAI, [S.l.], 2022. Disponível em: https://openai.com/index/chatgpt/. Acesso em: 03 dez. 2025.
OPENAI. OpenAI API. OpenAI, [S.l.], 2020. Disponível em: https://openai.com/index/openai-api/. Acesso em: 03 dez. 2025.
RENNISON, Callie Marie; HART, Thomas C. Research methods in criminal justice and criminology. Thousand Oaks: SAGE Publications, 2019.
SCHREIER, Margrit. Sampling and generalization. In: FLICK, Uwe. (Org.). The SAGE handbook of qualitative data collection. London: SAGE Publications, 2018, p. 84-98.
SOR, Jennifer. Sam Altman touts ChatGPT's 800 million weekly users, double all its main competitors combined. Business Insider, [S.l.], 2025. Disponível em: https://www.businessinsider.com/chatgpt-users-openai-sam-altman-devday-llm-artificial-intelligence-2025-10. Acesso em: 03 dez. 2025.
STRYKER, Cole et al. What is deep learning? In: Cole STRYKER et al. (eds.). The 2025 Guide to Machine Learning. [S.l.]: IBM, 2025. Disponível em: https://www.ibm.com/think/topics/deep-learning. Acesso em: 30 nov. 2025.
STRYKER, Cole; SCAPICCHIO, Mark. What is generative AI? In: Cole STRYKER et al. (eds.). The 2025 Guide to Machine Learning. [S.l.]: IBM, 2025. Disponível em: https://www.ibm.com/think/topics/generative-ai. Acesso em: 27 nov. 2025.
SUN, Dawei; BENSON, Michael L. Mixed methods: A justification, explication, and example. In: FARIA, Rita; DODGE, Mary. (Org.). Qualitative research in criminology: Cutting-edge methods. Cham: Springer, 2022. p. 37-49.
TREND MICRO RESEARCH. Malicious Uses and Abuses of Artificial Intelligence. Europol’s European Cybercrime Centre, 2020.
TWOMEY, John et al. What Is So Deep About Deepfakes? A Multi-Disciplinary Thematic Analysis of Academic Narratives About Deepfake Technology. IEEE Transactions on Technology and Society, v. 6, n. 1, p. 64-79, 2025.
UNIVERSIDADE DE HARVARD. The Benefits and Limitations of Generative AI: Harvard Experts Answer Your Questions. Harvard, 2023. Disponível em: https://harvardonline.harvard.edu/blog/benefits-limitations-generative-ai. Acesso em: 27 nov. 2025.
UNIVERSIDADE DE LEEDS. What is Generative AI? University of Leeds, Leeds, [S.d.]. Disponível em: https://generative-ai.leeds.ac.uk/intro-gen-ai/strengths-and-weaknesses/. Acesso em: 27 nov. 2025.
WAGNER, Travis L.; BLEWER, Ashley. The Word Real Is No Longer Real: Deepfakes, Gender, and the Challenges of AI-Altered Video. Open Information Science, v. 3, n. 1, 2019.
WHITTAKER, Lucas et al. Mapping the deepfake landscape for innovation: A multidisciplinary systematic review and future research agenda. Technovation, v. 125, p. 01-17. 2023.
WILES, Rose. What are qualitative research ethics? London: Bloomsbury Academic, 2013.
Downloads
Publicado
Como Citar
Edição
Seção
Licença
Copyright (c) 2025 Pedro Miguel Freitas, Ana Guerreiro

Este trabalho está licenciado sob uma licença Creative Commons Attribution-NonCommercial 4.0 International License.
Na qualidade de autor(es) da colaboração, original e inédita, sobre o qual me(nos) responsabilizo(amos) civil e penalmente pelo seu conteúdo, após ter lido as diretrizes para autores, concordado(amos) plenamente com as Políticas Editorias da Revista Novos Estudos Jurídicos - NEJ e autorizo(amos) a publicação na rede mundial de computadores (Internet), permitindo, também, que sua linguagem possa ser reformulada, caso seja necessário, sem que me(nos) seja devido qualquer pagamento a título de direitos autorais, podendo qualquer interessado acessá-lo e/ou reproduzi-lo mediante download, desde que a reprodução e/ou publicação obedeçam as normas da ABNT e tenham a finalidade exclusiva de uso por quem a consulta a título de divulgação da produção acadêmico científico.



















