Objectives:
This article delineates the changing landscape brought by implementing Artificial Intelligence (AI) in criminal case analysis, focusing on its potential role in minimizing the risk of wrongful convictions. The study assesses both the opportunities AI offers for improving adjudication quality and efficiency, and the significant barriers to its implementation, such as the need for full files digitalization and ensuring algorithmic transparency.

Methods:
The analysis examines the systemic challenge of wrongful convictions, drawing parallels to the DNA revolution as a historical precedent for technological change. It explores the use of Machine Learning (ML) and Large Language Models (LLMs), leveraging Natural Language Processing (NLP), to analyze vast legal documents for procedural errors, repetitive patterns, and latent biases that contribute to judicial failures. The study also incorporates empirical findings on AI-assisted trial efficiency and error reduction

Results:
AI can act as a strategic mechanism by detecting patterns and minimizing subjective biases, thereby complementing human judgment. In post-conviction review, AI can assist organizations in selecting cases with the highest potential for exoneration. However, barriers include the necessity of system standardization, high implementation costs, the risk of judicial dependence, and the "black box" problem hindering credibility.

Conclusions:
AI emerges as the next stage in the evolution towards procedural reliability, offering unprecedented data analysis capabilities to reduce judicial errors. While its potential is significant, AI is not yet a milestone in this fight. Effective implementation requires solid foundations: full digitalization, IT system standardization, and the creation of legal and ethical frameworks that guarantee transparency and retain essential human oversight.