#FactCheck- AI-Generated Deepfake Falsely Shows Lt Gen Devendra Sharma Speaking About Missing Aviators in Operation Sindoor
Executive Summary
A purported video of Lieutenant General Devendra Sharma is being shared on social media, falsely portraying him as saying that “five aviators who participated in Operation Sindoor are missing from our training list. CyberPeace Research Wingresearch found that the viral video is completely fabricated and generated using artificial intelligence. The Ministry of Defence has also dismissed the clip as an AI-generated deepfake. The investigation found that Lieutenant General Devendra Sharma, Commander of the Army Training Command, never made any such statement. The viral clip has been manipulated by inserting fake audio into footage from his original speech to spread a misleading narrative.
Claim
A user on social media platform X shared the viral video with the caption:“Fifty-nine officers were listed at the Combat Army Aviation Training School, but five officers are missing. They were involved in Operation Sindoor as pilots of four Rafale jets and one Su-30 aircraft…”
Fact Check
To verify the claim, we extracted several keyframes from the viral video and conducted a reverse-image search. During the investigation, we found the original video on ANI’s official X account. In the authentic footage, Lt Gen Sharma is seen speaking at a routine event at a training institution. His remarks focused on regular training-related matters, and he made no reference to missing pilots, Operation Sindoor, or any alleged losses.
https://x.com/ANI/status/2061731994971881667
In the next stage of the investigation, we found a post by the Ministry of Defence, Government of India, which also identified the viral clip as a deepfake.
https://www.facebook.com/watch/?v=1697698548109000
The ministry clarified that the video had been manipulated using AI-generated audio and did not reflect any genuine statement made by Lt Gen Devendra Sharma.
Conclusion
The viral video is a fabricated AI-generated deepfake. Lieutenant General Devendra Sharma never made the statement attributed to him about five missing aviators linked to Operation Sindoor. The original video was altered by inserting fake audio into authentic footage. The manipulated clip is being circulated online to spread misinformation and create confusion regarding the Indian Armed Forces.
Related Blogs
Somewhere in a compliance meeting right now, someone is saying "we have eighteen months, we're fine." That sentence is doing the same thing a snooze button does at 6 a.m.: technically buying time, while quietly making the actual wake up call worse. India's data protection law just started its countdown, and the 18 months everyone keeps citing is not a grace period to procrastinate through. It is closer to a runway before takeoff. Runways exist for one purpose: building up speed until the plane has no choice but to leave the ground. Standing still on one is not a strategy.
What actually got notified, and when
On 13 November 2025, the Ministry of Electronics and Information Technology notified the Digital Personal Data Protection Rules, 2025, giving operational shape to the Digital Personal Data Protection Act that Parliament had passed back in August 2023. Alongside the Rules themselves, MeitY issued a separate Enforcement Notification setting out exactly when different provisions kick in, and a further notification establishing the Data Protection Board of India, headquartered in the National Capital Region with four members. The final Rules followed a genuinely deliberative process, MeitY had floated draft Rules in January 2025 for public consultation and received 6,915 individual inputs from startups, industry bodies, civil society groups, and citizens before finalising the version now in force. The headline structural decision, and the one causing the most confusion in boardrooms, is that the Rules do not commence all at once. They commence in three distinct phases spread across eighteen months, and different obligations become legally binding at each stage.
The phased timeline, laid out plainly
That third date, 13 May 2027, is the one that matters most for the vast majority of organisations, since it is where the bulk of actual operational obligations, the parts that touch product design, customer facing notices, and breach response, become enforceable. Legal commentary tracking the rollout has been consistent that this is described as a hard deadline with no grace period expected once it arrives, since the Data Protection Board is already operational and can begin receiving complaints well before Phase 3 obligations formally take effect.
Why "later" is a genuinely expensive plan
The financial stakes attached to Phase 3 non-compliance are not modest. The Schedule to the DPDP Act sets fixed penalty ceilings rather than turnover linked fines, which sounds gentler than Europe's GDPR model until you look at the actual numbers. Failure to implement reasonable security safeguards that results in a data breach can draw a penalty of up to 250 crore rupees per instance, the single highest tier in the Schedule. Failing to notify the Board or affected individuals after a breach occurs can draw up to 200 crore rupees, as can non-compliance with the Act's specific protections for children's data. Because these are assessed per instance rather than as a single capped exposure, a single incident that trips more than one obligation, say, inadequate safeguards that also delay breach notification, can compound into penalty exposure running into hundreds of crores from one event. All penalties collected go to the Consolidated Fund of India rather than to affected individuals directly, meaning the deterrent is aimed squarely at organisational behaviour, not compensation.
The part everyone keeps underestimating: this is not just a legal department problem
Perhaps the most consequential shift buried inside the DPDP framework is who actually has to own it. Reading the Rules as a checklist for the legal or privacy team alone misses how far the obligations actually reach. Building a compliant consent lifecycle touches product design. Security safeguards touch cybersecurity and IT infrastructure directly. Retention and deletion logic touches data governance and engineering. Third party risk review touches procurement. Breach preparedness touches internal audit and incident response. And increasingly, as organisations deploy AI systems that process personal data, AI governance enters the picture too, since a model trained or fine tuned on personal data inherits the same DPDP obligations as any other processing activity.
That cross functional reality is where most readiness programmes currently fall short. Treating DPDP compliance as a documentation exercise, updating a privacy policy PDF and calling it done, produces the appearance of compliance without the operational substance a Data Protection Board investigation would actually test. A breach response plan that exists only on paper and has never been rehearsed will not hold up against the 72 hour data principal notification window the Rules impose once Phase 3 lands. A consent mechanism bolted onto a website without corresponding backend logic to honour withdrawal requests will not satisfy an actual audit.
What a serious readiness posture looks like right now
Organisations that are ahead of this curve are already treating the eighteen month window as three overlapping workstreams rather than one deadline to hit at the end.
- The first is discovery: mapping what personal data exists, where it flows, who owns each system that touches it, and why it is collected in the first place, since compliance is structurally impossible without first knowing what you are protecting. This stage typically surfaces uncomfortable findings, shadow data sets nobody formally owns, vendor integrations nobody fully mapped, legacy systems still holding data well past any reasonable retention justification.
- The second is build: standing up the actual mechanisms, consent flows that can genuinely honour a withdrawal request end to end, rights request handling that does not depend on a single overworked employee checking an inbox, retention and deletion logic wired into the systems themselves rather than described only in a policy document, and security controls proportionate to the sensitivity of what is being protected.
- The third is proof: generating the internal evidence, audit trails, documented decisions, tested response procedures, that demonstrates governance was real rather than retrofitted after the fact. A Data Protection Board investigation, when it eventually happens, will not be satisfied by a well written policy; it will look for evidence that the policy was actually operational.
The actual question worth asking
The right question was never "when does DPDP become enforceable." Phase 1 already answered that; the law is live, and the Data Protection Board already exists and can act. The better question, the one worth taking into any leadership review between now and May 2027, is simpler and considerably less comfortable: will the organisation actually be ready when each phase's obligations become operational, or will readiness be assembled in a scramble once the deadline stops being theoretical. 18 months sounds long right up until the week it does not, and by the time Phase 3 lands, "we'll get to it" will no longer be a sentence any organisation gets to finish.
References
- Ministry of Electronics and Information Technology, Digital Personal Data Protection Rules, 2025, notified 13 November 2025. Press Information Bureau, "Digital Personal Data Protection Rules, 2025 Notified," 14 November 2025. https://static.pib.gov.in/WriteReadData/specificdocs/documents/2025/nov/doc20251117695301.pdf
- Shardul Amarchand Mangaldas & Co, "Enforcement of the DPDP Act and notification of the DPDP rules." https://www.amsshardul.com/insight/enforcement-of-the-dpdp-act-and-notification-of-the-dpdp-rules/
- S&R Associates, "India's Digital Personal Data Protection Regime Takes Effect." https://www.snrlaw.in/indias-digital-personal-data-protection-regime-takes-effect/
- Khurana & Khurana, "MeitY Notifies Rules Operationalising The DPDP Framework." https://www.khuranaandkhurana.com/update-meity-notifies-rules-operationalising-the-dpdp-framework-in-india
- Exchange4media, "DPDP Act 2025: Penalties for violations can reach Rs 250 crore." https://www.exchange4media.com/digital-news/dpdp-act-2025-penalties-for-confirmed-violations-can-reach-rs-250-crore-149360.html
- Seclore, "DPDP Rules 2025: India's Complete Compliance Guide." https://www.seclore.com/fundamentals/dpdp-rules-2025-compliance-guide/
Introduction
On 1 July 2026, the United Nations published their first draft of the Development of Artificial Intelligence (AI) report. The scientists who authored this document are 40 individuals who comprise the Independent International Scientific Panel on Artificial Intelligence , which was formed by the UN General Assembly in 2025 and chaired by Bengio and Ressa.
The report’s key takeaway message relates to the importance of countries establishing frameworks for the appropriate use of AI technologies as a means of preventing the aggravation of social inequality. UN Secretary-General António Guterres pointed out that without such regulation, society will not be able to enjoy the benefits of these technologies. After all, their explosive development is already beginning to have negative consequences.
This report served as key for the UN Global Dialogue on AI Governance, which was held in Geneva on 6thJuly, 2026.
What Exactly Does the Report Say?
It is not a rulebook, it doesn't create laws. It is more of a survey to get countries on board before they start treaties regarding AI.
This is seen through how much AI has developed from previous technologies, such as electricity taking decades to be common everywhere while the internet took 15 years to get to a billion users, but chatGPT alone got to 100 million users within 2 months. Today, millions of people use AI chatbots, and AI has helped map out 200 million+ protein structures currently helping us get drugs to market much faster.
The report is also cautious. It talks about an "evidence dilemma" , we need evidence to make good AI rules, but by the time it's collected, it might be too late to control the technology. There are 45+ AI frameworks worldwide, but they're disconnected, and some safety checks are done by the same companies that build the AI, raising doubts about their trustworthiness.
The Big Inequality Problem
This is one of the most eye-opening parts of the report. It says the USA has about 75% of the world's most powerful supercomputers, while China has around 15%, together, 90% of the world's top computing power. Most AI models are also built in these two countries, meaning they stand to gain the most from AI's growth.
Here's the shocking part: nearly two-thirds of people in high-income countries will benefit from AI tools. But in low-income countries, only about 5% will see any real benefit. Just because AI tools exist doesn't mean everyone gets equal access — especially poorer countries that depend on other nations for AI tools, cloud services, and data.
The Language Problem
Another major issue is language. Most AI tools work well in English and a few major languages. But for other languages, these tools are far less reliable, which becomes dangerous in fields like healthcare.
The report provides shocking examples of how AI translation software has been mistakenly conflating smallpox with syphilis and gonorrhoea with diabetes. It also describes how an attempt by a nurse to reassure a patient speaking Tigrinya that she was about to be given "IV antibiotics", had in fact, been translated by the software into "IV insecticides".
Other Serious Concerns
The report also warns about darker uses of AI, like creating sexual abuse material and fake explicit videos (deepfakes), which mostly harm women and children. AI-generated disinformation is being used to manipulate elections, while cybercriminals use AI for fraud and other crimes.
In addition, AI could have negative impacts on individuals struggling with mental health issues and even driving vulnerable populations to commit suicide. Moreover, the global AI infrastructure is estimated to release huge amounts of CO2 that is contributing to environmental damage.
What the Report Recommends
Because most of the power over AI development will be controlled by major corporations and nations, the report charges that the system is likely to lead to "unjust hierarchical control over critical systems."
To fix this, the suggestions include:
- Independent groups, not AI companies themselves, should test AI systems.
- Countries should cooperate to regulate artificial intelligence technologies instead of imposing a policy on a particular state.
- The states should develop the technical and language expertise to participate in the global negotiations on AI governance.
- However, the global AI governance should address developing countries’ needs to upgrade their infrastructures, such as electricity and data, and strike a balance between sustainability and digitalization.
Conclusion
This UN report is an important first step, giving countries real scientific evidence as they move towards binding international agreements on AI. Its biggest takeaway is this: AI itself is neutral — neither good nor bad on its own. What makes it good or bad for the world are the choices made by governments, firms, and societies in regulating it. Without proper control, AI technologies may well make inequality even worse than it already is. The negotiations to come will be critical in determining the future of the technology.
References
- United Nations News, 'AI explained: Why the world needs to act now', 1 July 2026
- Euronews, 'Window to control AI is closing and it could widen inequality, UN experts warn', 2 July 2026
- Irish Examiner, 'Rapid spread of AI may worsen global inequality, UN warns', 1 July 2026
- Reuters, 'UN report sees enormous potential benefits, big risks in AI', 1 July 2026
- Preliminary Report of the Independent International Scientific Panel on Artificial Intelligence, United Nations, July 2026
Executive Summary
A video featuring Union Defence Minister Rajnath Singh is being widely circulated on social media with the claim that he issued a strict warning against student protesters. In the viral clip, Singh is allegedly heard saying that the government would not tolerate “anarchy” and that strict action would be taken against those blocking roads, disturbing public order, or damaging government property. CyberPeace Research Wing’s research found the claim to be false. The research revealed that the original video available on Rajnath Singh’s official YouTube channel was digitally altered by adding AI-generated audio. The statement heard in the viral video is not genuine and was created using artificial intelligence.
Claim:
An account on social media platform X (formerly Twitter), with the username @ForumDefence, shared a video on July 22, 2026, claiming that Union Defence Minister Rajnath Singh had issued a warning regarding the ongoing student protest. The video features an AI-generated voice attributed to Rajnath Singh, in which he is allegedly heard saying: “Taking advantage of the student movement, the unrest and violence being spread over the past few days are strongly condemned by the government. Expressing one’s views is a democratic right, but breaking the law and creating difficulties for the public in its name will not be tolerated. The government will not tolerate such anarchy. Anyone attempting to block roads, disturb peace, or damage government property will face strict action.”
The link, archive link and screenshot of the post are provided below:
https://x.com/ForumDefence/status/2079960247888756775
Fact Check:
To verify the viral claim, the Desk conducted a keyword search on Google. However, no credible news report or official statement was found confirming that Rajnath Singh had made such remarks regarding the student protest.
During further research, the Desk found the original video on the official YouTube channel of Union Defence Minister Rajnath Singh. A comparison between the viral clip and the original video showed that the statements attributed to Singh in the viral video did not match his actual remarks. In the original video, Rajnath Singh was speaking about the ongoing protests in Delhi. He said that the protests were a matter of serious concern and expressed that it was unfortunate that some people were attempting to use students and youth as a political tool for their own interests.
He further stated that the government was committed to addressing the concerns of students and youth, and that dialogue and discussion were the appropriate ways to resolve issues in a democracy.
The original video from Rajnath Singh’s official YouTube channel can be accessed here:
https://www.youtube.com/watch?v=w73iPTEfKCA
Further analysis of the viral video’s audio was conducted using the AI detection tool Resemble AI. The analysis indicated that the audio had a 99.10 per cent probability of being AI-generated.
The Desk also examined the viral video using the AI detection tool HIVE Moderation. The analysis suggested that the speech used in the video had an 88.1 per cent probability of being AI-generated.
Conclusion:
The research established that the viral video featuring Union Defence Minister Rajnath Singh has been digitally manipulated by adding AI-generated audio. The statement attributed to Singh regarding strict action against student protesters is not genuine. The original video available on Rajnath Singh’s official YouTube channel contains different remarks, and the viral clip has been circulated with a false and misleading claim.
