#FactCheck: Fake viral AI video captures a real-time bridge failure incident in Bihar

Introduction

Artificial intelligence is often hailed as a democratiser of knowledge, opportunity and skill. It is set to improve diagnostics, personalised learning, and productivity to boost the economy, which can assist millions of people to leave poverty. However, this may be an incomplete picture. A report of the United Nations Development Programme in 2025 tells a more complex tale. The Next Great Divergence: Why AI May Widen Inequality Between Countries cautions that, unless acts are taken to intervene, AI will not alleviate inequality between countries but will instead concentrate benefits in already advantaged economies and increase risks in more vulnerable ones.

Two Gaps, One Crisis

AI is not going to create a level playing field: it has been injected into a world where there is unprecedented inequality. The report outlines two structural asymmetries that will influence the ways in which its effects manifest: a capability gap and a vulnerability gap. 

Those countries that have high connectivity, skills, compute and regulation will be in a position to reap a greater portion of the AI dividend. Others will be exposed to greater risks of job losses, information exclusion, misinformation, and the indirect consequences of increased energy and water demands. 

The centre of this transition is the Asia-Pacific region, that harbors a population of more than 55 per cent of the world. More than half of the global AI users are now located in the region, but the initial positions are quite different. Nations such as Singapore and South Korea are already spending a lot of money on AI infrastructure, with others still striving to offer basic broadband services. Two out of three individuals already use AI tools in certain high-income economies. In most countries with low incomes, the utilisation is lower. Such figures are important as they depict not only a gap in technology but also a structural difference in terms of who controls AI and who is controlled by the latter.

When Inequality Becomes a Trust Problem

Any trusted technological system is based on three tenets: transparency, fairness and accountability. AI inequality negatively impacts all three.

If governments implement imported AI systems in areas with limited technical capability, with limited transparency on their operation, their construction, and their biases. Citizens do not really trust when decision-making systems are black boxes and domestic institutions lack the know-how to question them.

Data exclusion also interferes with fairness. The AI systems trained with the datasets not sufficiently representative of the rural population, linguistic minorities, and women will generate poorer results in those groups systematically. Since South Asian women are much less likely to own a smartphone,  this impacts their representation in digital data, and consequently in any AI system trained on such data. 

Safety Risks Are Not Evenly Distributed

The lack of trust has a direct safety aspect. For example, those countries that have less robust information ecosystems have a greater exposure to AI-generated misinformation that can bias the discourse of the populace, alter elections, and cause violence. They also have the weakest capability of screening, tagging, or combating such content.

The same can be said about labour markets. The very same technologies that can speed up marginalisation and destabilise governance increase human insecurity, especially among employees in the informal economy with weak social security. The UNDP report points out that the exposure of female employment to disruption by AI is disproportionate to that of male employment, which further presents a gendered dimension in an already unequal situation.

Risks of infrastructure are skewed as well. Large AI systems may create disproportionately high energy and water demands on countries that host the data infrastructure without there being an equivalent economic payback. The environmental cost is local while profits are outsourced. Dangers of AI spread downwards, and the advantages go upwards.

The Governance Gap and Regulatory Arbitrage

Governance is perhaps the most important aspect. There are only a few states that presently have extensive AI regulation systems. This gives rise to a patchy landscape, in which safety standards differ dramatically and where companies have an incentive to install systems in jurisdictions that have weaker regulation.

The main reason is the lack of capability, as expressed by Philip Schellekens, chief economist of the UNDP in Asia and the Pacific, who says that those countries that invest in skills, computing power and well-run governance structures will gain. The rest will be left far behind. 

This departure has its ramifications outside the nations. When users in other areas are subjected to widely different rates of safety and equity by the same international platforms, the concept of uniform digital norms would no longer be sustainable. Confidence in AI systems is lost not only locally but also on a global scale.

Way Forward

The UNDP report makes it clear that there is no inevitability of divergence. To avert it, however, it is necessary to consider AI governance as a development, rather than a technology problem.

The capacity to govern should be constructed and not presumed. This implies assisting countries in establishing regulatory systems, institutional capacity, and facilitating cross-border collaboration on standards. It can also imply considering some AI features as a public good, with common models and open standards that do not allow a few firms or states to become too powerful.

The UNDP articulates the problem in a simple manner: in the end, the world's people and not machines must decide on what technologies should be given priority and how to utilise them optimally.

Conclusion

AI inequality is often framed as an economic divergence story. But its implications run deeper. It reshapes who is protected, who is visible in data, and who has the power to challenge harmful outcomes. The risk is not just that some countries fall behind economically. It is that the global digital ecosystem fragments into zones of high trust and low trust, high protection and low protection. The choices made now will determine which path prevails. AI can reinforce existing divides or help bridge them. 

But that outcome will not be decided by the technology itself. It will be decided by how societies choose to distribute access, power, and responsibility in the systems they build.

References

1. https://www.undp.org/sites/g/files/zskgke326/files/2025-12/undp-rbap-the-next-great-divergence_1.pdf
2. https://www.undp.org/asia-pacific/press-releases/ai-risks-sparking-new-era-divergence-development-gaps-between-countries-widen-undp-report-finds
3. https://www.undp.org/asia-pacific/blog/next-great-divergence-how-ai-could-split-world-again-if-we-dont-intervene
4. https://www.aljazeera.com/news/2025/12/2/ai-threatens-to-widen-inequality-among-states-un
5. https://www.undp.org/asia-pacific/next-great-divergence
6. https://www.eco-business.com/press-releases/ai-risks-spark-new-era-of-divergence-as-development-gaps-widen-undp-report/

‍

Introduction

Search Engine Optimisation (SEO) is a process through which one can improve website visibility on search engine platforms like Google, Microsoft Bing, etc. There is an implicit understanding that SEO suggestions or the links that are generated on top are the more popular information sources and, hence, are deemed to be more trustworthy. This trust, however, is being misused by threat actors through a process called SEO poisoning.

SEO poisoning is a method used by threat actors to attack and obtain information about the user by using manipulative methods that position their desired link, web page, etc to appear at the top of the search engine algorithm. The end goal is to lure the user into clicking and downloading their malware, presented in the garb of legitimate marketing or even as a valid result for Google search.

An active example of attempts at SEO poisoning has been discussed in a report by the Hindustan Times on 11th November, 2024. It highlights that using certain keywords could make a user more susceptible to hacking. Hackers are now targeting people who enter specific words or specific combinations in search engines. According to the report, users who looked up and clicked on links at the top related to the search query “Are Bengal cats legal in Australia?” had details regarding their personal information posted online soon after.

SEO Poisoning - Modus Operandi Of Attack

There are certain tactics that are used by the attackers on SEO poisoning, these are:

• Keyword stuffing- This method involves overloading a webpage with irrelevant words, which helps the false website appear higher in ranking.
• Typosquatting- This method involves creating domain names or links similar to the more popular and trusted websites. A lack of scrutiny before clicking would lead the user to download malware, from what they thought was a legitimate site.
• Cloaking- This method operates by showing different content to both the search engines and the user. While the search engine sees what it assumes to be a legitimate website, the user is exposed to harmful content.
• Private Link Networks- Threat actors create a group of unrelated websites in order to increase the number of referral links, which enables them to rank higher on search engine platforms.
• Article Spinning- This method involves imitating content from other pre-existing, legitimate websites, while making a few minor changes, giving the impression to search engine crawlers of it being original content.
• Sneaky Redirect- This method redirects the users to malicious websites (without their knowledge) instead of the ones the user had intended to click.

CyberPeace Recommendations

• Employee Security Awareness Training: Security awareness training can help employees familiarise themselves with tactics of SEO poisoning, encouraging them to either spot such inconsistencies early on or even alert the security team at the earliest. 
• Tool usage: Companies can use Digital Risk Monitoring tools to catch instances of typosquatting. Endpoint Detection and Response (EDR) tools also help keep an eye on client history and assess user activities during security breaches to figure out the source of the affected file.
• Internal Security Measures: To refer to lists of Indicators of Compromise (IOC). IOC has URL lists that show evidence of the strange behaviour of websites, and this can be used to practice caution. Deploying Web Application Firewalls (WAFs) to mitigate and detect malicious traffic is helpful.

Conclusion

The nature of SEO poisoning is such that it inherently promotes the spread of misinformation, and facilitates cyberattacks. Misinformation regarding the legitimacy of the links and the content they display, in order to lure users into clicking on them, puts personal information under threat. As people trust their favoured search engines, and there is a lack of awareness of such tactics in use, one must exercise caution while clicking on links that seem to be popular, despite them being hosted by trusted search engines.

References

• https://www.checkpoint.com/cyber-hub/cyber-security/what-is-cyber-attack/what-is-seo-poisoning/
• https://www.vectra.ai/topics/seo-poisoning
• https://www.techtarget.com/whatis/definition/search-poisoning
• https://www.blackberry.com/us/en/solutions/endpoint-security/ransomware-protection/seo-poisoning
• https://www.coalitioninc.com/blog/seo-poisoning-attacks
• https://www.sciencedirect.com/science/article/abs/pii/S0160791X24000186
• https://www.repindia.com/blog/secure-your-organisation-from-seo-poisoning-and-malvertising-threats/ 
• https://www.hindustantimes.com/technology/typing-these-6-words-on-google-could-make-you-a-target-for-hackers-101731286153415.html

‍

Amid reports that the death toll in Iran’s ongoing protests has risen to 2,571, a video has been widely circulated on social media showing a man slapping a person dressed in clerical attire after an argument. Users sharing the clip claim that public anger in Iran has escalated to the point where people are now physically attacking religious clerics. However, research by the Cyber Peace Foundation has found this claim to be misleading. The research  established that the video is not recent and has no connection to the current protests in Iran. In fact, the clip dates back to 2021 and was entirely scripted.

‍

Claim

‍

On January 14, 2026, users on X (formerly Twitter) shared the viral video with captions suggesting that Iranian citizens are openly assaulting clerics amid the ongoing unrest. One such post stated that the situation in Iran had deteriorated so badly that people were now beating religious leaders.

‍

The link, archived version, and screenshot of the post are available below:

• https://x.com/brijeshchaodhry/status/2011316139423605203 
• https://archive.ph/yclOD 

‍

‍

Factcheck:

To verify the authenticity of the claim, the Cyber Peace Foundation extracted keyframes from the viral video and conducted a Google reverse image search. This led investigators to a report published on April 19, 2021, on the Persian-language website of Deutsche Welle (DW). The visuals matched the viral clip exactly, confirming that the footage is nearly five years old, not recent. Here is the link to the original video, along with a screenshot:

• https://www.dw.com/fa-ir/%D9%82%D8%AF%D8%B1%D8%AA-%D9%88-%D8%B3%D8%B1%D8%B9%D8%AA-%D8%B9%D9%85%D9%84-%D9%BE%D9%84%DB%8C%D8%B3-%D8%AF%D8%B1-%D8%AF%D8%B3%D8%AA%DA%AF%DB%8C%D8%B1%DB%8C-%D8%B9%D9%88%D8%A7%D9%85%D9%84-%DA%A9%D9%84%DB%8C%D9%BE-%D8%B3%DB%8C%D9%84%DB%8C-%D8%A8%D9%87-%DB%8C%DA%A9-%D8%B1%D9%88%D8%AD%D8%A7%D9%86%DB%8C/a-57249985

‍

Further examination of reports by Fars News Agency revealed that Tehran police had conducted a detailed probe into the video at the time and declared it fake and pre-scripted. According to Tehran Police Chief Hossein Rahimi, the individual seen wearing religious attire was not a cleric. Here is the link to the original video, along with a screenshot: He was actually employed at a carpet cleaning shop in Tehran, while the man seen slapping him was his own son.

Police stated that the video was deliberately staged and circulated to provoke public sentiment and create unrest by falsely linking it to religious tensions. Both the father and son were arrested, and images of them in police custody were published in contemporaneous reports. Additional confirmation was found on the Independent Persian website, which had also reported on the incident on April 19, 2021, reiterating that the video was fabricated and unrelated to any protest movement. Here is the link to the original video, along with a screenshot:

•  https://www.independentpersian.com/node/139891/%D8%B3%DB%8C%D8%A7%D8%B3%DB%8C-%D9%88-%D8%A7%D8%AC%D8%AA%D9%85%D8%A7%D8%B9%DB%8C/%D9%BE%D9%84%DB%8C%D8%B3-%D9%85%DB%8C%E2%80%8C%DA%AF%D9%88%DB%8C%D8%AF%E2%80%8C-%D9%81%DB%8C%D9%84%D9%85-%C2%AB%D8%B3%DB%8C%D9%84%DB%8C-%D8%AE%D9%88%D8%B1%D8%AF%D9%86-%D8%A2%D8%AE%D9%88%D9%86%D8%AF%C2%BB-%D8%B3%D8%A7%D8%AE%D8%AA%DA%AF%DB%8C-%D8%A8%D9%88%D8%AF-%D9%88-%D8%B3%D8%A7%D8%B2%D9%86%D8%AF%DA%AF%D8%A7%D9%86-%D8%A8%D9%87-%D8%AF%D9%86%D8%A8%D8%A7%D9%84

‍

‍

Conclusion

The claim that the viral video shows an Iranian protester slapping a cleric during the current wave of protests is false. The video is from 2021, was scripted, and has no link to the ongoing demonstrations in Iran. It is being reshared with a misleading narrative to spread disinformation and inflame public sentiment.c

‍