#FactCheck- Viral Zipline Accident Video from Karnaprayag is AI-Generated, Claim is False
Executive Summary
A video is being shared on social media claiming that a female tourist from Delhi fell into a deep gorge during ziplining in Karnaprayag. In the viral video, a woman dressed in bright pink and yellow clothes is seen ziplining amidst high mountains, when suddenly the zipline wire breaks and she falls from a significant height into a deep gorge. Social media users are sharing this video, presenting it as a real incident. Research by the CyberPeace Research Wing revealed that the claim of a Delhi woman tourist's death during a zipline accident in Karnaprayag is baseless. The viral video is AI-generated.
Claim
According to the claim, this accident took place in Karnaprayag, Uttarakhand, where a female tourist from Delhi became a victim of this horrific mishap. Several social media users are claiming that the woman died in this accident.
https://www.facebook.com/reel/26108544868822432
Fact Check
To investigate the video viral as a zipline accident in Karnaprayag, we conducted a reverse search of its keyframes. During this, we found the video uploaded as a Short on a Pakistani YouTube channel, @Zoyaqueen-w2t, on June 2, 2026. Hashtags like '#funnyshorts' were used in the caption of this video. Scanning this account revealed that several other fictional videos related to similar zipline accidents have also been uploaded here. Apart from this, the video was also shared with funny hashtags in some social media posts from May 2026.
https://www.youtube.com/shorts/Gzha_J7Fqv0
Following this, searching with relevant keywords yielded no credible media reports regarding any such zipline accident in Karnaprayag, Uttarakhand, in recent times. Subsequently, we scanned the viral video of the alleged zipline accident using the AI detection tool 'Hive Moderation'. During the analysis, the tool classified the video as highly likely to be AI-generated, with a score of 99.2%.
Conclusion
From the evidence gathered in our research , it is clear that the claim regarding the death of a female tourist from Delhi during a zipline accident in Karnaprayag is baseless. The viral video is not of a real incident, but is AI-generated.
Related Blogs
Introduction
The rise in start-up culture, increasing investments, and technological breakthroughs are being encouraged alongside innovations and the incorporation of generative Artificial Intelligence elements. Witnessing the growing focus on human-centred AI, its potential to transform industries like education remains undeniable. Enhancing experiences and inculcating new ways of learning, there is much to be explored. Recently, a Delhi-based non-profit called Rocket Learning, in collaboration with Google.org, launched Appu- a personalised AI educational tool providing a multilingual and conversational learning experience for kids between 3 and 6.
AI Appu
Developed in 6 months, along with the help of dedicated Google.org fellows, interactive Appu has resonated with those the founders call “super-users,” i.e. parents and caregivers. Instead of redirecting students to standard content and instructional videos, it operates on the idea of conversational learning, one equally important for children in the targeted age bracket. Designed in the form of an elephant, Appu is supposed to be a personalised tutor, helping both children and parents understand concepts through dialogue. AI enables the generation of different explanations in case of doubt, aiding in understanding. If children were to answer in mixed languages instead of one complete sentence in a single language (eg, Hindi and English), the AI would still consider it as a response. The AI lessons are two minutes long and are inculcated with real-world examples. The emphasis on interactive and fun learning of concepts through innovation enhances the learning experience. Currently only available in Hindi, it is being worked on to include 20 other languages such as Punjabi and Marathi.
UNESCO, AI, and Education
It is important to note that such innovations also find encouragement in UNESCO’s mandate as AI in education contributes to achieving the 2030 Agenda of Sustainable Development Goals (here; SDG 4- focusing on quality education). Within the ambit of the Beijing Consensus held in 2019, UNESCO encourages a human-centred approach to AI, and has also developed the “Artificial Intelligence and Education: Guidance for Policymakers” aiming towards understanding its potential and opportunities in education as well as the core competencies it needs to work on. Another publication was launched during one of the flagship events of UNESCO- (Digital Learning Week, 2024) - AI competency frameworks for both, students and teachers which provide a roadmap for assessing the potential and risks of AI, each covering common aspects such as AI ethics, and human-centred mindset and even certain distinct options such as AI system design for students and AI pedagogy for teachers.
Potential Challenges
While AI holds immense promise in education, innovation with regard to learning is contentious as several risks must be carefully managed. Depending on the innovation, AI’s struggle with multitasking beyond the classroom, such as administrative duties and tedious grading, which require highly detailed role descriptions could prove to be a challenge. This can become exhausting for developers managing innovative AI systems, as they would have to fit various responses owing to the inherent nature of AI needing to be trained to produce output. Security concerns are another major issue, as data breaches could compromise sensitive student information. Implementation costs also present challenges, as access to AI-driven tools depends on financial resources. Furthermore, AI-driven personalised learning, while beneficial, may inadvertently reduce student motivation, also compromising students' soft skills, such as teamwork and communication, which are crucial for real-world success. These risks highlight the need for a balanced approach to AI integration in education.
Conclusion
Innovations related to education, especially the ones that focus on a human-centred AI approach, have immense potential in not only enhancing learning experiences but also reshaping how knowledge is accessed, understood, and applied. Untapped potential using other services is also encouraged in this sector. However, maintaining a balance between fostering intrigue and ensuring the inculcation of ethical and secure AI remains imperative.
References
- https://www.unesco.org/en/articles/what-you-need-know-about-unescos-new-ai-competency-frameworks-students-and-teachers?hub=32618
- https://www.unesco.org/en/digital-education/artificial-intelligence
- https://www.deccanherald.com/technology/google-backed-rocket-learning-launches-appu-an-ai-powered-tutor-for-kids-3455078
- https://indianexpress.com/article/technology/artificial-intelligence/how-this-google-backed-ai-tool-is-reshaping-education-appu-9896391/
- https://www.thehindu.com/business/ai-appu-to-tutor-children-in-india/article69354145.ece
- https://www.velvetech.com/blog/ai-in-education-risks-and-concerns/
Introduction
Artificial Intelligence (AI) is fast transforming our future in the digital world, transforming healthcare, finance, education, and cybersecurity. But alongside this technology, bad actors are also weaponising it. More and more, state-sponsored cyber actors are misusing AI tools such as ChatGPT and other generative models to automate disinformation, enable cyberattacks, and speed up social engineering operations. This write-up explores why and how AI, in the form of large language models (LLMs), is being exploited in cyber operations associated with adversarial states, and the necessity for international vigilance, regulation, and AI safety guidelines.
The Shift: AI as a Cyber Weapon
State-sponsored threat actors are misusing tools such as ChatGPT to turbocharge their cyber arsenal.
- Phishing Campaigns using AI- Generative AI allows for highly convincing and grammatically correct phishing emails. Unlike the shoddily written scams of yesteryears, these AI-based messages are tailored according to the victim's location, language, and professional background, increasing the attack success rate considerably. Example: It has recently been reported by OpenAI and Microsoft that Russian and North Korean APTs have employed LLMs to create customised phishing baits and malware obfuscation notes.
- Malware Obfuscation and Script Generation- Big Language Models (LLMs) such as ChatGPT may be used by cyber attackers to help write, debug, and camouflage malicious scripts. While the majority of AI instruments contain safety mechanisms to guard against abuse, threat actors often exploit "jailbreaking" to evade these protections. Once such constraints are lifted, the model can be utilised to develop polymorphic malware that alters its code composition to avoid detection. It can also be used to obfuscate PowerShell or Python scripts to render them difficult for conventional antivirus software to identify. Also, LLMs have been employed to propose techniques for backdoor installation, additional facilitating stealthy access to hijacked systems.
- Disinformation and Narrative Manipulation
State-sponsored cyber actors are increasingly employing AI to scale up and automate disinformation operations, especially on election, protest, and geopolitical dispute days. With LLMs' assistance, these actors can create massive amounts of ersatz news stories, deepfake interview transcripts, imitation social media posts, and bogus public remarks on online forums and petitions. The localisation of content makes this strategy especially perilous, as messages are written with cultural and linguistic specificity, making them credible and more difficult to detect. The ultimate aim is to seed societal unrest, manipulate public sentiments, and erode faith in democratic institutions.
Disrupting Malicious Uses of AI – OpenAI Report (June 2025)
OpenAI released a comprehensive threat intelligence report called "Disrupting Malicious Uses of AI" and the “Staying ahead of threat actors in the age of AI”, which outlined how state-affiliated actors had been testing and misusing its language models for malicious intent. The report named few advanced persistent threat (APT) groups, each attributed to particular nation-states. OpenAI highlighted that the threat actors used the models mostly for enhancing linguistic quality, generating social engineering content, and expanding operations. Significantly, the report mentioned that the tools were not utilized to produce malware, but rather to support preparatory and communicative phases of larger cyber operations.
AI Jailbreaking: Dodging Safety Measures
One of the largest worries is how malicious users can "jailbreak" AI models, misleading them into generating banned content using adversarial input. Some methods employed are:
- Roleplay: Simulating the AI being a professional criminal advisor
- Obfuscation: Concealing requests with code or jargon
- Language Switching: Proposing sensitive inquiries in less frequently moderated languages
- Prompt Injection: Lacing dangerous requests within innocent-appearing questions
These methods have enabled attackers to bypass moderation tools, transforming otherwise moral tools into cybercrime instruments.
Conclusion
As AI generations evolve and become more accessible, its application by state-sponsored cyber actors is unprecedentedly threatening global cybersecurity. The distinction between nation-state intelligence collection and cybercrime is eroding, with AI serving as a multiplier of adversarial campaigns. AI tools such as ChatGPT, which were created for benevolent purposes, can be targeted to multiply phishing, propaganda, and social engineering attacks. The cross-border governance, ethical development practices, and cyber hygiene practices need to be encouraged. AI needs to be shaped not only by innovation but by responsibility.
References
- https://www.microsoft.com/en-us/security/blog/2024/02/14/staying-ahead-of-threat-actors-in-the-age-of-ai/
- https://www.bankinfosecurity.com/openais-chatgpt-hit-nation-state-hackers-a-28640
- https://oecd.ai/en/incidents/2025-06-13-b5e9
- https://www.microsoft.com/en-us/security/security-insider/meet-the-experts/emerging-AI-tactics-in-use-by-threat-actors
- https://www.wired.com/story/youre-not-ready-for-ai-hacker-agents/
- https://www.cert-in.org.in/PDF/Digital_Threat_Report_2024.pdf
- https://cdn.openai.com/threat-intelligence-reports/5f73af09-a3a3-4a55-992e-069237681620/disrupting-malicious-uses-of-ai-june-2025.pdf
.webp)
Introduction
India's Competition Commission of India (CCI) on 18th November 2024 imposed a ₹213 crore penalty on Meta for abusing its dominant position in internet-based messaging through WhatsApp and online display advertising. The CCI order is passed against abuse of dominance by the Meta and relates to WhatsApp’s 2021 Privacy Policy. The CCI considers Meta a dominant player in internet-based messaging through WhatsApp and also in online display advertising. WhatsApp's 2021 privacy policy update undermined users' ability to opt out of getting their data shared with the group's social media platform Facebook. The CCI directed WhatsApp not to share user data collected on its platform with other Meta companies or products for advertising purposes for five years.
CCI Contentions
The regulator contended that for purposes other than advertising, WhatsApp's policy should include a detailed explanation of the user data shared with other Meta group companies or products specifying the purpose. The regulator also stated that sharing user data collected on WhatsApp with other Meta companies or products for purposes other than providing WhatsApp services should not be a condition for users to access WhatsApp services in India. CCI order is significant as it upholds user consent as a key principle in the functioning of social media giants, similar to the measures taken by some other markets.
Meta’s Stance
WhatsApp parent company Meta has expressed its disagreement with the Competition Commission of India's(CCI) decision to impose a Rs 213 crore penalty on them over users' privacy concerns. Meta clarified that the 2021 update did not change the privacy of people's personal messages and was offered as a choice for users at the time. It also ensured no one would have their accounts deleted or lose functionality of the WhatsApp service because of this update.
Meta clarified that the update was about introducing optional business features on WhatsApp and providing further transparency about how they collect data. The company stated that WhatsApp has been incredibly valuable to people and businesses, enabling organization's and government institutions to deliver citizen services through COVID and beyond and supporting small businesses, all of which further the Indian economy. Meta plans to find a path forward that allows them to continue providing the experiences that "people and businesses have come to expect" from them. The CCI issued cease-and-desist directions and directed Meta and WhatsApp to implement certain behavioral remedies within a defined timeline.
The competition watchdog noted that WhatsApp's 2021 policy update made it mandatory for users to accept the new terms, including data sharing with Meta, and removed the earlier option to opt-out, categorized as an "unfair condition" under the Competition Act. It was further noted that WhatsApp’s sharing of users’ business transaction information with Meta gave the group entities an unfair advantage over competing platforms.
CyberPeace Outlook
The 2021 policy update by WhatsApp mandated data sharing with Meta's other companies group, removing the opt-out option and compelling users to accept the terms to continue using the platform. This policy undermined user autonomy and was deemed as an abuse of Meta's dominant market position, violating Section 4(2)(a)(i) of the Competition Act, as noted by CCI.
The CCI’s ruling requires WhatsApp to offer all users in India, including those who had accepted the 2021 update, the ability to manage their data-sharing preferences through a clear and prominent opt-out option within the app. This decision underscores the importance of user choice, informed consent, and transparency in digital data policies.
By addressing the coercive nature of the policy, the CCI ruling establishes a significant legal precedent for safeguarding user privacy and promoting fair competition. It highlights the growing acknowledgement of privacy as a fundamental right and reinforces the accountability of tech giants to respect user autonomy and market fairness. The directive mandates that data sharing within the Meta ecosystem must be based on user consent, with the option to decline such sharing without losing access to essential services.
References
