Facial Recognition Evidence: Legal and Ethical Limits

Share on:

Introduction

With the increasing adoption of Artificial Intelligence (AI) in law enforcement, Facial Recognition Technology (FRT) has emerged as a powerful tool for identifying suspects, tracking movements, and verifying identities. In India, FRT is already in use by various state police departments and central agencies. While it holds the potential to revolutionize criminal investigations, its evidentiary value in court raises critical questions about due process, accuracy, bias, and the ethical use of AI forensics.

This article explores the legal and ethical limits of facial recognition evidence, focusing on its admissibility in Indian courts, concerns around reliability, and the need for procedural safeguards.

Understanding Facial Recognition Evidence

Facial Recognition Technology uses AI algorithms to analyze images or video footage and match facial features against a database. When used in investigations, it can potentially identify suspects from CCTV footage, social media, or public surveillance networks.

In India, systems like the National Automated Facial Recognition System (AFRS) have been proposed to integrate data across various law enforcement databases. The Delhi Police has reportedly used facial recognition tools during protests and public events, raising questions of transparency and accountability.

Evidentiary Use

Facial recognition can be used in two broad ways:

• Investigative Lead: It helps police identify persons of interest or narrow down suspects. At this stage, it functions as an aid and not as conclusive proof.
• Evidentiary Proof: FRT outputs are used in court to establish identity or presence at a scene. This raises questions about whether such evidence meets the standards under the Indian Evidence Act, 1872.

Legal Framework in India

Indian Evidence Act, 1872

Under Section 3 of the Indian Evidence Act, evidence includes "all documents, including electronic records." Section 65B allows for the ‘admissibility of electronic records,’ provided certain conditions are met regarding authenticity and certification.

Facial recognition outputs would be considered electronic records, but courts must evaluate whether:

• The system used is reliable and scientifically valid
• The match is statistically significant
• The evidence is not tampered with or biased

Due Process and Constitutional Rights

The use of FRT must conform to Article 21 of the Constitution, which guarantees the right to life and personal liberty, including the right to privacy (as upheld in Justice K.S. Puttaswamy v. Union of India, 2017). The judgment mandates that any intrusion into privacy must meet the tests of:

• Legality: Backed by a valid law
• Necessity: Pursues a legitimate aim
• Proportionality: Least intrusive means
• Procedural Safeguards

Currently, India lacks a comprehensive data protection law, which makes the unchecked use of facial recognition potentially unconstitutional.

Ethical and Accuracy Concerns

Accuracy and Bias

One of the primary concerns with facial recognition evidence is algorithmic accuracy, especially in varied lighting, angles, and crowd conditions. Studies show:

• Higher false positive rates for darker-skinned individuals and women
• Lower accuracy when images are of poor quality
• Risk of misidentification leading to wrongful arrests

The National Institute of Standards and Technology (NIST), USA, found that even top facial recognition systems exhibited racial and gender bias.

Black Box Problem

FRT algorithms often function as "black boxes"; their decision-making process is opaque. This raises issues when a defendant wishes to challenge the methodology or confidence level of the match.

In a criminal trial, where “proof beyond reasonable doubt” is the standard, reliance on a non-transparent system without the ability to cross-examine or audit the process violates due process.

International Standards and Comparative Jurisprudence

United States

In People v. Lopez, 2020, a New York court held that FRT results alone could not establish probable cause for arrest. Courts in the U.S. have increasingly demanded algorithmic transparency and expert validation before accepting FRT evidence.

European Union

The General Data Protection Regulation (GDPR) and proposed AI Act strictly regulate biometric data. In most EU jurisdictions, real-time facial recognition in public spaces is banned or heavily restricted.

Lessons for India

India should draw on international practices to frame its legal approach to FRT, particularly around:

• Consent and notice requirements
• Accuracy benchmarks
• Auditability of AI systems
• Exclusionary rules for flawed or biased evidence

Safeguards for Admissibility in Indian Courts

To ensure that facial recognition evidence does not violate constitutional rights or lead to miscarriage of justice, Indian law should adopt the following safeguards:

Technical Validation

• Use only NIST-certified or independently tested algorithms
• Maintain confidence score thresholds for match acceptance
• Allow defendants access to metadata, match scores, and logs

Procedural Safeguards

• FRT should not be the sole basis for conviction
• Independent expert testimony should accompany the evidence
• Ensure chain of custody for digital records
• Allow cross-examination of the FRT process and operator

Legal Reforms

• Enact a Facial Recognition Regulation Law covering permissible use, retention limits, and audit mechanisms
• Amend the Indian Evidence Act to include standards for AI-generated evidence
• Include judicial oversight for deploying FRT in sensitive scenarios (e.g., protests, political gatherings)

The Future of AI Forensics in India

As India moves toward integrating AI forensics into criminal justice, the use of facial recognition must be approached with caution. Technologies like gait analysis, emotion detection, and voice recognition are also emerging, but without proper regulation, they risk infringing on basic rights.

The challenge lies in balancing technological efficiency with individual liberty. While FRT can reduce investigation time and enhance accuracy in identifying repeat offenders, it must not bypass the fundamental principles of justice.

Conclusion

Facial Recognition Technology, when used responsibly, can be a useful tool in criminal justice. However, its use as facial recognition evidence must be subject to stringent accuracy checks, legal scrutiny, and due process safeguards. Given the lack of a dedicated legal framework in India, courts must adopt a cautious approach, ensuring that such evidence meets high standards of reliability and respects constitutional rights.

The right to a fair trial and the presumption of innocence are cornerstones of our legal system. Any evidence, especially one generated by AI forensics in India, must be tested not only for relevance but also for fairness. Without robust checks, the promise of FRT could turn into a threat to civil liberties.