Challenges and opportunities of cybersecurity-oriented AI in the legal system
TL;DR
The literature review indicates that AI can improve cybersecurity threat detection and legal information processing but also introduces risks involving inaccurate outputs, malicious use, manipulated information, transparency, and accountability, which support continued development of AI systems that are secure, transparent, and accountable.
Abstract
Artificial intelligence (AI) is increasingly being used in cybersecurity and legal systems to process information, identify patterns, and support decision-making. As it becomes more integrated into high-stakes environments, its reliability and security have become important concerns. AI can improve cybersecurity through applications such as threat detection and intrusion monitoring, while also creating new vulnerabilities through phishing, social engineering, and other malicious uses. In legal environments, AI can assist with legal research and document analysis, but studies have identified problems involving inaccurate legal information, fabricated citations, algorithmic bias, and limited transparency. These concerns are further complicated by AI-generated misinformation and the need to protect the integrity of digital information used in legal processes. However, fewer studies have examined how cybersecurity vulnerabilities and AI-generated misinformation may jointly influence trust in AI-assisted legal environments. This study reviews literature at the intersection of AI, cybersecurity, and legal systems and develops Legal Risk Analyzer (https://github.com/chloesalgado5/legal-risk-analyzer), a Python programing language-based document analyzer as a proof-of-concept screening tool. The literature review indicates that AI can improve cybersecurity threat detection and legal information processing but also introduces risks involving inaccurate outputs, malicious use, manipulated information, transparency, and accountability. Furthermore, a brief demonstration of the custom document analyzer has differentiated four constructed sample passages according to predefined indicators, assigning different reliability and cybersecurity risk scores based on citations, uncertainty language, absolute language, and cybersecurity-related terms. Yet, it suggests that rule-based screening alone cannot determine whether a legal claim is factually accurate. Taken together, as AI becomes increasingly integrated into legal and cybersecurity systems, maintaining public trust will require these areas to be considered together rather than independently. Combining cybersecurity protections with legal verification and responsible AI governance may help organizations identify risks before AI-generated information influences high-stakes decisions. These findings support continued development of AI systems that are secure, transparent, and accountable.