Is Bharatiya Sakshya Adhiniyam Ready for Synthetic Evidence? — Truth Becomes a Technical Question in the Age of AI Deepfakes?

This Blog is Written by Vidushi Singh, 4th Year, BBA LLB (H), Amity University, Ranchi, Jharkhand.

Blog 01 | Edition VIII

When Seeing Doesn't Mean Believing

Suppose that a video is shown to a judge. It depicts the accused giving a confession. The voice is clear, the voice is not fake, and the file hasn't been changed since it was acquired. But the accused made no confession. The entire video was created with AI. It's not a thing that can be done. It's an obvious problem of evidence. Synthetic media and deepfakes are capable of creating seemingly real pictures, sounds and events. The problem with the law is that it must ask itself when a digital record appears true, but the reality shown is not real, what does it do?

The Evolution of Digital Evidence Into Synthetic Evidence

Litigation has undergone a revolution in the form of electronic evidence. The facts can be proven by CCTV footage, mobile recordings, emails, photographs and cloud records as they would be based on testimony alone. In line with its requirements for evidentiary purpose, the Bharatiya Sakshya Adhiniyam, 2023 (BSA) specifically acknowledges the use of electronic or digital records. Generative AI poses another challenge, however. Traditional manipulation is a modification of an original recording. Synthetic evidence can produce the evidence without any real world event. A fake voice can sound like a real voice; AI-generated video can depict a person doing something that they never actually did. The difference is important because a file may be an authentic digital object, but a false representation of the real life.

What is the BSA Testing for?

The statutory regulation of electronic and digital records is contained in sections 61-63 of the BSA. Section 61 prohibits the rejection of an electronic or digital record being rejected solely because it is electronic, while Section 62 and 63 relate to the admissibility of and proof of such records. The Schedule to Section 63 also considers more technical aspects, including the electronic record source device and the hash value of an electronic record. (India Code) The protections are for integrity and provenance, and deepfakes present a conceptual challenge. A hash can tell you that a specific file has not been altered, but it can't tell you that the event in the hash actually occurred. “Is this the same file?” and “Is this a true description of reality?” are thus separate questions. The mere formality of being technically accurate is not enough to establish the facts.

The 2026 IT Rules: A Regulatory Warning

India has already taken steps to solve the problem apart from the court of law. The amendments in the Information Technology Rules came into effect from 20 February 2026, introducing a new term: “synthetically generated information” (SGI). The framework also applies to realistic AI generated or manipulated AV content, and stipulates that there be metadata attached or another system of documentation to provide a source for the synthetic content, if technically possible. It also imposes duties on intermediaries with regards to illegal synthetic content. This is important for evidence law as it acknowledges the possibility of a content appearing real, but actually being artificially-generated. However, the role of platforms in regulation and proof in court are different. Disputed material must be admitted and relied upon in a trial, which does not depend on a label.

The Authentication Trap

The main issue can be put simply as:

Authenticity ≠ Accuracy ≠ Truth.

Let's say one video is downloaded from a video platform that has been created by artificial intelligence. The downloaded copy has been authentic. Its hash is intact. It has an adequate chain of custody. But the incident in it is completely fictional. This is an original file. Content is man-made. The event is a fictional one. That will effectively push future evidentiary practice past the point of asking if a digital file has been altered. It will also have to assess the source and process of the content it presents and, when challenged, verify if there is independent evidence that the event it depicts actually took place.

Is There a Technological Problem that Can be Solved by Technology?

Courts can be helped by digital forensics, metadata analysis, cryptographic hashes, watermarking, and provenance systems and AI-based detectors. None should be relied upon as infallible, however. True evidence may be misinterpreted and elaborate fake evidence may go undetected. The judicial decision-making process should therefore be aided by expert analysis, rather than be supplanted by it. Digital evidence should be used in conjunction with other evidence and bearing in mind independent corroboration, if applicable.

Who Needs to Prove It?

When one party makes a video and the other states that it's a “deepfake”, the law needs to answer the question of how to handle it. The burden should not be such that the actual digital evidence becomes impractical to use. Meanwhile, complicated synthetic documentation is not to be accepted as authentic simply because it says that it is. A reasonable approach would be to have better checks on high-risk audio-visual evidence in the presence of a credible challenge, beginning with proving the provenance and basic integrity, and then, if appropriate, with proper forensic examination of it by the producing party.

The Constitutional Stakes

The implications are not limited to the evidence law. A fake recording can have a detrimental impact on reputation, privacy and liberty. Use of synthetic evidence in criminal cases may compromise the right to a fair trial and lead to the conviction of an innocent person. The Supreme Court recently warned of a danger from fake legal authorities created by AI; a similar threat is posed to the very process of adjudication when AI creates fake material as if it were genuine material. In Pooja Ramesh Singh vs. Jammu and Kashmir Bank Ltd, decided on July 2026, the Court quashed orders which were based on non-existent or AI-generated authorities, and emphasized the need for verification. It is not necessary for India to have its own “deepfake evidence law.” It requires a framework that is clear, and can separate file integrity from factual authenticity. Courts could provide more precise rules for competing synthetic evidence, provide more specialized digital-forensic services, enhance preservation and chain of custody, and consider evidence of provenance when technically sound or feasible. Meanwhile, there needs to be protection from misleading accusations of a "deepfake" being legitimate evidence.

Conclusion

AI has not only the problem of being able to manipulate evidence, but it can also be used to generate evidence that appears convincing. That’s when the truth itself can be created in a manner that seems like evidence. The BSA has established an important framework for electronic and digital records, but synthetic media puts the assumptions to the test in a new way. The courtroom can not simply ask the question, “Has this file been altered?” It must also pose the question of what fundamental nature is. Was this reality what was represented in this file? Truth doesn’t always have to be “what the evidence proves”, in the era of AI. The question could very well be at first a technical one: do we believe that evidence?

 

(Write to the author at vidushisingh281104@gmail.com.)