The Use of Artificial Intelligence in Combating War Crimes: Ethical and Legal Challenges

The increasing digitization of armed conflicts has positioned artificial intelligence (AI) as a transformative tool in war crime investigations, offering unprecedented capabilities in evidence collection, perpetrator identification, and atrocity prevention. This study examines AI's dual role as both an evidentiary asset and a source of ethical-legal challenges within international criminal justice systems. Through doctrinal analysis and empirical case studies (Syria, Myanmar, Ethiopia), we demonstrate how machine learning enhances geospatial intelligence (85% more incidents documented than traditional methods) and biometric recognition, while also revealing systemic risks: racial bias in facial recognition (31% error rates for darker-skinned individuals) and evidentiary opacity incompatible with ICC standards (Article 69(4), Rome Statute). The research identifies three critical gaps: (1) normative ambiguities in AI-generated evidence admissibility (Prosecutor v. Bemba, 2016), (2) accountability vacuums for algorithmic errors, and (3) privacyviolative data practices (34,000 civilians exposed in Sudan, 2023). We propose a four-pillar regulatory framework: (1) International AI Evidence Standards mandating blockchain-based provenance tracking, (2) Algorithmic Transparency Protocols requiring open-source audits under EU AI Act Article 29, (3) War Crime AI Literacy Programs for legal professionals (currently only 32% of ICC prosecutors understand ML basics), and (4) Rome Statute Reforms establishing an AI Evidence Review Chamber. Cross-referencing 24 jurisdictions, the study demonstrates that human rights-compliant AI deployment requires "explainability by design" (Wachter, 2017) and layered oversight—particularly for high-risk applications like predictive policing (44% false positives in Palestine). The findings advocate for a balanced paradigm where AI augments, but never replaces, human legal judgment in upholding in dubio pro reo principles.