The Threat of “Deepfakes” in International War Crime Investigations

Abstract

Digital evidence has become fundamental in modern war crime investigations, specifically in conflicts where investigators face limited physical access. At the same time, the increase of deepfakes—highly realistic videos or audio recordings produced using AI—creates a major threat to the reliability of such material. This article evaluates the challenges deepfakes create for accountability in the International Criminal Court (ICC). It discusses the legal framework under the Rome Statute, relevant ICC case law, and the growing reliance on digital evidence in situations such as Syria, Myanmar, Ukraine, and Gaza. The article argues that deepfakes undermine authenticity, disrupt chain of custody standards, and weaken public trust in the evidence used in court. It concludes with practical proposals for improving verification processes through stronger guidelines and technological security measures.

Introduction

International criminal investigations increasingly depend on digital assets, including videos, photos, audio recordings, satellite imagery, and open-source intelligence (OSINT). In conflicts such as those in Ukraine, Syria, Myanmar, and Gaza, these sources play a central role in documenting alleged war crimes. The ICC’s Office of the Prosecutor has confirmed that digital submissions now form a significant part of the evidence collected for ongoing investigations, reflecting the reality that investigators are often unable to access active conflict zones physically.

However, the rise of deepfake technology has introduced serious new risks. Deepfakes utilize machine-learning techniques to create convincing videos or audio clips depicting events or statements that never occurred. Their constant growth and sophistication make them difficult to detect, even with professional forensic tools. In the context of war crime investigations—where authenticity and reliability are crucial—deepfakes threaten to distort facts, mislead investigators, and undermine fair trial rights.

This article assesses the legal and practical challenges created by deepfakes, explains why they are difficult to address under current evidentiary standards, and proposes methods to strengthen the reliability of digital evidence in international criminal accountability.

Legal Framework: Evidentiary Standards under the Rome Statute

The ICC’s evidentiary rules prioritize three core tenets: authenticity, relevance, and probative value. Article 69(4) of the Rome Statute instructs the Court to consider the reliability of evidence, including any prejudice to a fair trial or a fair evaluation of the testimony of a witness. Furthermore, Article 74 requires that judgments rely only on evidence that has been scrutinized by both parties. Deepfakes directly threaten these principles because they appear highly realistic while being entirely fabricated, thus challenging the Court’s ability to determine probative value.

Relevant ICC Jurisprudence

Although the ICC has not yet adjudicated a case hinging specifically on deepfake evidence, existing case law highlights the Court’s cautious approach to digital evidence:

1. Prosecutor v. Lubanga (2012): The Trial Chamber warned that evidence collected through intermediaries requires careful verification due to the inherent risk of manipulation.

2. Prosecutor v. Al-Mahdi (2016): The Court relied on videos showing the destruction of cultural heritage but insisted on corroboration through multiple sources, including expert analysis.

3. Prosecutor v. Ongwen (2021): The Chamber emphasized that digital evidence must be authenticated through rigorous metadata analysis and supporting testimonial evidence.

These cases illustrate that while the ICC takes digital authenticity seriously, deepfakes create sophisticated challenges that exceed the risks anticipated in earlier jurisprudence.

Critical Analysis

How Deepfakes Distort Investigations

Deepfakes possess the potential to fabricate war crimes that never occurred, falsely depict commanders giving unlawful orders, or create fake admissions of guilt. Beyond direct fabrication, they can also undermine genuine evidence by creating a “liar’s dividend”—casting doubt on authentic footage by suggesting it could be fake.

Chain of Custody Challenges

Digital evidence often lacks a clear chain of custody. Videos are frequently uploaded, downloaded, and shared across multiple platforms, a process that can strip essential metadata and obscure the file’s origin. Deepfakes exploit this weakness, allowing bad actors to insert fabricated material into the information ecosystem without an identifiable creator.

Impact on Fair Trial Rights

A convincing deepfake could wrongfully incriminate an innocent person or allow a guilty party to attack the credibility of genuine inculpatory evidence. Both outcomes undermine fair trial rights and weaken public confidence in international tribunals.

Difficulty of Detection

Deepfake technology is evolving faster than available detection tools. While forensic methods such as pixel analysis, shadow examination, and audio waveform testing are useful, advanced deepfakes are increasingly capable of evading these standard techniques.

Recent Developments

Technological Responses

Civil society organizations play a critical role in verification. Groups like Bellingcat, Amnesty International’s Evidence Lab, and Human Rights Watch utilize geolocation, chronolocation, metadata extraction, and satellite verification to authenticate media. While these methods strengthen the verification of materials brought before the courts, they require expert handling and do not guarantee absolute accuracy.

Institutional Responses

UN investigative bodies have developed secure digital upload systems to better preserve metadata. Similarly, the ICC’s Technology Office is expanding its capacity to analyze digital files, including the integration of AI-based detection tools.

Proposal

To effectively combat the threat of deepfakes, the international legal community should adopt the following measures:

1. Expand Technological Capacity: The ICC must invest in specialized digital forensic teams and advanced tools to accurately detect AI-driven content.

2. Education and Awareness: Investigators and the public must be educated on the limitations of digital content to limit the spread of misinformation.

3. Develop ICC-Specific Guidelines: The Court should adopt clear, codified rules for the authentication and detection of synthetic media.

4. Improve Chain-of-Custody Practices: Secure evidence upload platforms should be standardized and encouraged for use by NGOs and civilians to preserve original metadata.

5. Promote International Cooperation: The ICC should collaborate with tech companies and forensic laboratories to establish global verification standards.

Conclusion

Deepfakes represent one of the most serious challenges to accountability in international criminal law. They not only threaten innocent parties with wrongful persecution but also cast doubt on authentic evidence essential for justice. The ICC and the UN must take proactive steps to ensure the verification of digital evidence. Only by securing the reliability of evidence can the international justice system preserve fair trial rights, support victims, and maintain its legitimacy.

CHANGE LOG

• Structure Consolidation: Merged multiple bulleted lists (specifically in the Introduction and Critical Analysis) into prose. Why: Excessive bullet points disrupt the flow of reading and are less professional for a legal article.

• Grammar & Phrasing Corrections:

  • Changed “growing reliability” to “growing reliance.” Why: “Reliability” refers to trustworthiness; “reliance” refers to dependence.

  • Changed “At the same” to “At the same time.” Why: Corrects a missing word in a common idiom.

  • Changed “100% fake” to “entirely fabricated.” Why: “100% fake” is too colloquial for a legal context.

  • Corrected “This article asses” to “This article assesses.”

• Header Refinement: Renamed “Statutory Interpretation” to “Relevant ICC Jurisprudence.” Why: The section discussed case law (Lubanga, Al-Mahdi), not the interpretation of statutory text.

• Citation Formatting: Italicized case names (Prosecutor v. Lubanga) and standard legal terms. Why: Adheres to standard legal citation conventions (Bluebook/OSCOLA standards).

• Sentence Repair: Fixed the fragment in the Introduction (“Within the context…”). Why: To ensure grammatical correctness and clarity.

• Voice Preservation: Retained the specific case examples and the structure of the “Proposal” section. Why: To honor the author’s research and distinct concluding recommendations.

RECOMMENDATIONS FOR FUTURE SUBMISSIONS

1. In-Text Citations: Legal writing requires that claims be supported immediately in the text. Do not rely solely on a bibliography. Use footnotes or in-text references (e.g., “Article 69(4) of the Rome Statute”) to validate your arguments.

2. Minimize Bullet Points: Use bullet points sparingly, primarily for distinct recommendations or lengthy lists. For standard lists (like types of evidence), use prose to maintain a professional narrative flow.

3. Strengthen Transitions: Ensure that each section flows logically into the next. Use transition sentences to connect your arguments (e.g., moving from the risks of deepfakes to the legal precedents that might address them).

4. Precise Legal Terminology: Be careful with terms like “Statutory Interpretation” versus “Case Law.” Ensure headers accurately reflect the content beneath them.

5. Proofread for Fragments: Read your work aloud to catch sentence fragments and run-on sentences, particularly in the introduction and analysis sections.