Authored By: Soumil Roy
IILM university
INTRODUCTION
For centuries, our legal system has proven to be remarkably adaptable, as it evolves throughout the ages with every major technological revolution. From the birth of the printing press to the dawn of the current digital age, the law has consistently recalibrated its standards of proof to accommodate new dimensions of evidence while keeping our legal system fair and just. Yet, we have arrived at a juncture where the judiciary faces a variable that may be the most difficult to counter in our legal history: hyper-realistic, AI-generated video. Unlike previous innovations the more it advances the more it changes the basis of how much a court can now trust any eye witness testimony during a trail as AI video has reached a state where we can’t trust our own eyes without putting someone’s freedom on the line
In the context of the Indian legal system, this threat becomes more urgent as we move further into this new era of technology. The last time a major act was made was as recently as 2023 with the Bharatiya Sakshya Adhiniyam [1]. But even though the act was passed recently, technology is not sleeping; it is advancing every day, and many people are going to abuse it. Our legal system has to be ready to take them on. While the system has been trying its best, the age of hyper-realistic AI is coming closer and closer. The further the legal system delays taking action, the further the danger of the freedom of others being lost to the “fake truths” made by AI.
Now this article examines the crisis by first detailing the existing legal framework in Section II and analyzing why hyper realistic AI break these traditional rules in Section III. Thereafter, Section IV will look at the global regulatory trends leading to Section V which proposes essential reforms to protect integrity of the Indian courts
THE EXISTING LEGAL FRAMEWORK
Recognition of Electronic Records under the BSA 2023
To properly understand why AI is such a different variable in the world of law, for over a century, India relied upon the Indian Evidence Act 1872, which was designed for a world of physical documents and paper trails However, as technology advanced, our legal system showed its adaptability by introducing the Bharatiya Sakshya Adhiniyam 2023 (BSA).[2] [3]Under section 61 of this new Act, the definition of a “document” includes “electronic and digital records,” meaning that any digital file—whether it is an email, a photo, or an AI clip—is now legally recognized as a document that can be used in a court of law.
The Gatekeeper: Section 63 and the Mandatory Certificate
While the law recognizes digital files in the court, it does not trust them automatically. Because since digital media is so easy to edit, and as per section 63 of the BSA acts as a “Gatekeeper.” For any electronic record to be allowed in court, it must be accompanied by a Section 63 Certificate. This is a signed document from a person in charge of the device stating that the hardware was working properly and that the data has not been tampered with.
The issue here is that the law focuses almost entirely on the hardware. It assumes that if the “machine” is working fine, the “image” it produced must be the truth. But now things have changed with hyper-realistic AI as it can create a “perfect” video on a perfectly functioning phone, which would make the evidence in front of us in video form harder to trust.
The Erosion of Oral and Documentary Evidence
Historically, the Indian legal system has relied on two pillars of proof: documentary evidence (physical papers) and oral evidence (witness testimony). Under the old rules, if a document’s authenticity was doubted, the court would rely on a witness to “prove” the contents through their own eyes and memory. However, now with the rise of hyper-realistic AI creates a “double-blind” situation. As AI videos get better, they don’t just mimic documents; they manipulate the content of what a witness sees. For example, a deep fake video can show a witness saying something they never said, or appearing in a place they never were. This breaks the traditional link between seeing a person and believing their testimony. Even with the Bharatiya Sakshya Adhiniyam 2023 trying to bridge the gap, the law still treats a witness’s “visual presence” on a screen as reliable. If the legal system cannot distinguish between a real person on camera and an AI-generated one, the very basis of “cross-examination” starts to fall apart.
III. THE EVOLUTION OF DECEPTION: FROM DARKROOMS TO ALGORITHMS
The Era of Manual Manipulation and Physical Artifacts
Now, in the world of law, misleading a jury and the court with fake visual evidence is not a modern occurrence. For as long as we have had the “silent witness” of photography, there have been those who tried to silence the truth through trickery. In the early days of the legal system, “doctoring” a photograph was a physical labor that happened in a darkroom. People used techniques like combination printing—literally layering two negatives on top of each other mismatched grain, ghostly edges where two images were joined, or shadows that pointed in two different directions. Because the law back then was built for a physical world, a judge could rely on these manual glitches to toss out a fake. The fault was in the human hand, and that hand always left a trace since humans are not perfect there was always an error that could have been found
The Digital Transition and the “Photoshop” Barrier
As we moved into the 1990s to 2000s, manipulation moved from the darkroom to the desktop. Software like Adobe Photoshop which allow for pixel-level editing. This was the first major test for the legal system’s adaptability, leading to the creation of the Section 63 (formerly 65B) Certificate system to verify the computer source.
Even though digital editing was faster, it still left digital footprints. Experts could detect a “Photo shopped” image by analyzing “Error Level Analysis” (ELA)—which looks for inconsistencies in how different parts of a JPEG are compressed[4]. If a face was pasted into a scene, it would have a different “noise” level than the background. Furthermore, a human editor still had to align lighting and reflections manually; a single mistake in a shadow’s direction was the “fault” that exposed the lie.
The Generative Leap: When “Faults” Disappear
What makes the current era of hyper-realistic AI so dangerous is that we have moved from simply editing reality to generating it from scratch. As with these algorithms advances, they remove the human “faults” that we used to rely on to spot a lie. When this technology was first being made during the era of 2022 to 2023, it had obvious problems that acted as a safety net for the law. These early deep fakes had visible glitches, such as people having six fingers, blurred hair textures, or unnatural eye blinking that looked “robotic” to a jury. Lawyers could easily point to these errors to prove a video was fake by watching it more than one time. However, as we move further into 2026, those safety nets are gone. Modern AI has advanced to a stage where it understands the physics of light and human movement perfectly. We are reaching a point where there is no mismatched shadow or “uncanny” movement to find because the AI calculates the entire scene as a single, flawless mathematical unit. When the faults that experts used for a century to catch liars finally vanish, the law’s current tools—and the most dangerous thought is that AI videos not going to stop advancing even after this.
COMPARATIVE PERSPECTIVE: GLOBAL RESPONSES TO THE AI THREAT
Now as we move into 2026 its becomes obvious that India is not alone in its struggle against AI videos as other global powers have been even more proactive in creating specific laws to counter the abuse of hyper-realistic AI.
The European Union: The AI Act and Mandatory Labelling
The European Union has taken the lead by passing the Artificial Intelligence Act 2024, the world’s first comprehensive legal framework for AI [5]. Unlike India’s “hardware-centric” approach, the EU focuses on transparency and accountability. As Under Article 50 of the AI Act, providers and developers of generative AI systems—especially those creating deep fakes—must ensure that their outputs are clearly labelled as “artificially generated” in a machine-readable format[6]. This means the technology itself is forced to “confess” its nature. By the time these rules become fully applicable in August 2026, the EU will have a system where a judge doesn’t have to guess; the digital file carries a permanent “warning label” that protects the court from being misled [7].
The United States: Digital Provenance and Safety Standards
In the United States, the approach has been more focused on “Digital Provenance”—tracking the history of a file from the moment it is created. Following a landmark Executive Order in 2023, the US government began pushing for new standards to verify the authenticity of government-issued content and to protect Americans from AI-driven fraud[8]. They are moving toward a system where only files with a “verified lineage” can be trusted as evidence. This is a major shift from our current Indian system, which only asks if a phone was “working properly” under Section 63 of the BSA. The US model suggests that in an era of “fake visuals,” we must stop trying to catch the fakes and instead focus on only trusting what we can prove is real
CRITICAL ANALYSIS AND RECOMMENDATIONS
Mandatory Cryptographic Watermarking and Provenance
The first step toward securing our legal system is moving away from visible labels and toward “Invisible Cryptographic Watermarking.” Currently, AI companies use watermarks that can be easily cropped out or edited. India should mandate a system where AI-generated content carries a non-removable, digital “DNA” embedded into the file’s metadata at the moment of creation. This would not limit the use of AI, but it would ensure that the video is always “searchable” by forensic tools. As seen in the Rashmika Mandanna [9]incident, the damage was done because the video circulated without any proof of its origin. If mandatory watermarking were the law, any social media platform or court official could have instantly verified it was a synthetic file, protecting the individual’s dignity without banning the use of AI videos itself.
B.”Safe Zones” for Public and Political Figures
To protect our democracy and the judicial process, the law must establish “High-Sensitivity Safe Zones” for the likenesses of political and judicial figures. We should implement a strict precaution that prevents AI models from generating hyper-realistic content of public officials without a verifiable digital signature from the official themselves. The 2024 Election deep fakes[10] showed that even a few hours of a “fake visual” can cause irreversible damage. By requiring AI developers to bake “safety filters” into their software that block the unauthorized use of public figures, we can prevent the legal system from being overwhelmed by political disinformation while still allowing the general public to use AI for creative and productive purposes.
CONCLUSION
So in the end, the rise of hyper-realistic AI video represents more than just a small new variable for a court, but rather a difficult challenge that hits the visual truth upon which the Indian judicial system has rested for over a century. While the legal system did transition from the Indian Evidence Act of 1872 to the Bharatiya Sakshya Adhiniyam (BSA) 2023 as it was a necessary step toward modernization, this article has demonstrated that procedural safeguards like the Section 63 certificate are increasingly inadequate. These tools were designed to verify by targeting the integrity of the hardware but not the authenticity of a reality generated by an algorithm. As the manual faults that once allowed human experts to detect deception—such as mismatched lighting or anatomical glitches—evaporate, the law faces a double-blind crisis where neither the witness nor the judge can trust their own senses.
So in these advancements of AI for the future our legal system has to advance with it as we must recognize that in the age of generative AI, the Right to a Fair Trial under Article 21 of the Constitution is now inextricably linked to the Right to Visual Truth. The recommendations proposed, including High-Sensitivity Safe Zones for public figures and the integration of digital provenance into the BSA, are not merely for technical updates; instead they are essential safeguards for a human liberty, privacy and security. The legal system must not wait for a catastrophic miscarriage of justice to occur before it recalibrates its standards of proof. By taking proactive legislative action today, the Indian judiciary can ensure that the silent witness of video evidence remains a tool for truth, rather than letting law breaks abuse such a power. The law must evolve at the speed of the algorithm, ensuring that even in an era of fake visual world, the sanctity of the courtroom remains fair for all and that the trust in the justice system remains absolute.
Reference(S):
[1] Bharatiya Sakshya Adhiniyam 2023.
[2] Bharatiya Sakshya Adhiniyam 2023, s 61.
[3] Bharatiya Sakshya Adhiniyam 2023, s 63.
[4] See D Raković, ‘Error Level Analysis (ELA)’ (2023) 78 Tehnika 445.
[5] Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonized rules on artificial intelligence (Artificial Intelligence Act) [2024] OJ L1689.
[6] Artificial Intelligence Act 2024, art 50.
[7] Artificial Intelligence Act 2024.
[8] Executive Order 14110 on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence (30 October 2023)
[9] Zeba Khan, ‘Rashmika Mandanna’s Viral Deepfake Video Will Leave You Aghast, Amitabh Bachchan Asks for Legal Action’ (WION News, 5 November 2023) <https://www.wionews.com/entertainment/watch-rashmika-mandannas-viral-deepfake-video-will-leave-you-aghast-amitabh-bachchan-asks-for-legal-action-655694\> accessed 16 April 2026
[10] Brennan Center for Justice, ‘Gauging the AI Threat to Free and Fair Elections’ (Brennan Center for Justice, 6 March 2025) <https://www.brennancenter.org/our-work/analysis-opinion/gauging-ai-threat-free-and-fair-elections\> accessed 16 April 2026
