#FactCheck-Fake Video of Mass Cheating at UPSC Exam Circulates Online

AI-generated content has been taking up space in the ever-changing dynamics of today's tech landscape. Generative AI has emerged as a powerful tool that has enabled the creation of hyper-realistic audio, video, and images. While advantageous, this ability has some downsides, too, particularly in content authenticity and manipulation. 

The impact of this content is varied in the areas of ethical, psychological and social harms seen in the past couple of years. A major concern is the creation of non-consensual explicit content, including nudes. This content includes content where an individual’s face gets superimposed onto explicit images or videos without their consent. This is not just a violation of privacy for individuals, and can have humongous consequences for their professional and personal lives. This blog examines the existing laws and whether they are equipped to deal with the challenges that this content poses. 

Understanding the Deepfake Technology

Deepfake technology is a media file (image, video, or speech) that typically represents a human subject that is altered deceptively using deep neural networks (DNNs). It is used to alter a person’s identity, and it usually takes the form of a “face swap” where the identity of a source subject is transferred onto a destination subject. The destination’s facial expressions and head movements remain the same, but the appearance in the video is that of the source. In the case of videos, the identities can be substituted by way of replacement or reenactment.

This superimposed content creates realistic content, such as fake nudes. Presently, creating a deepfake is not a costly endeavour. It requires a Graphics Processing Unit (GPU), software that is free, open-source, and easy to download, and graphics editing and audio-dubbing skills.  Some of the common apps to create deepfakes are DeepFaceLab and FaceSwap, which are both public and open source and are supported by thousands of users who actively participate in the evolution and development of these software and models. 

Legal Gaps and Challenges

Multiple gaps and challenges exist in the legal space for deepfakes and their regulation. They are:

1. The inadequate definitions governing AI-generated explicit content often lead to enforcement challenges.
2. Jurisdictional challenges due to the cross-border nature of crimes and the difficulties caused by international cooperation measures are in the early stages for AI content. 
3. There is a gap between the current consent-based and harassment laws for AI-generated nudes. 
4. Providing evidence or providing proof for the intent and identification of perpetrators in digital crimes is a challenge that is yet to be overcome. 

Policy Responses and Global Trends 

Presently, the global response to deepfakes is developing. The UK has developed the Online Safety Bill, the EU has the AI Act, the US has some federal laws such as the National AI Initiative Act of 2020 and India is currently developing the India AI Act as the specific legislation dealing with AI and its correlating issues. 

The IT Rules, 2021, and the DPDP Act, 2023, regulate digital platforms by mandating content governance, privacy policies, grievance redressal, and compliance with removal orders. Emphasising intermediary liability and safe harbour protections, these laws play a crucial role in tackling harmful content like AI-generated nudes, while the DPDP Act focuses on safeguarding privacy and personal data rights.

Bridging the Gap: CyberPeace Recommendations 

1. Initiate legislative reforms by advocating for clear and precise definitions for the consent frameworks and instituting high penalties for AI-based offences, particularly those which are aimed at sexually explicit material.
2. Advocate for global cooperation and collaborations by setting up international standards and bilateral and multilateral treaties that address the cross-border nature of these offences.
3. Platforms should push for accountability by pushing for stricter platform responsibility for the detection and removal of harmful AI-generated content. Platforms should introduce strong screening mechanisms to counter the huge influx of harmful content.  
4. Public campaigns which spread awareness and educate users about their rights and the resources available to them in case such an act takes place with them. 

Conclusion 

The rapid advancement of AI-generated explicit content demands immediate and decisive action. As this technology evolves, the gaps in existing legal frameworks become increasingly apparent, leaving individuals vulnerable to profound privacy violations and societal harm. Addressing this challenge requires adaptive, forward-thinking legislation that prioritises individual safety while fostering technological progress. Collaborative policymaking is essential and requires uniting governments, tech platforms, and civil society to develop globally harmonised standards. By striking a balance between innovation and societal well-being, we can ensure that the digital age is not only transformative but also secure and respectful of human dignity. Let’s act now to create a safer future!

References

• https://etedge-insights.com/technology/artificial-intelligence/deepfakes-and-the-future-of-digital-security-are-we-ready/
• https://odsc.medium.com/the-rise-of-deepfakes-understanding-the-challenges-and-opportunities-7724efb0d981 ‍
• https://insights.sei.cmu.edu/blog/how-easy-is-it-to-make-and-detect-a-deepfake/

Executive Summary:

The viral social media posts circulating several photos of Indian Army soldiers eating their lunch in the extremely hot weather near the border area in Barmer/ Jaisalmer, Rajasthan, have been detected as AI generated and proven to be false. The images contain various faults such as missing shadows, distorted hand positioning and misrepresentation of the Indian flag and soldiers body features. The various AI generated tools were also used to validate the same. Before sharing any pictures in social media, it is necessary to validate the originality to avoid misinformation.

‍

‍

Claims:

The photographs of Indian Army soldiers having their lunch in extreme high temperatures at the border area near to the district of Barmer/Jaisalmer, Rajasthan have been circulated through social media.

‍

Fact Check:

Upon the study of the given images, it can be observed that the images have a lot of similar anomalies that are usually found in any AI generated image. The abnormalities are lack of accuracy in the body features of the soldiers, the national flag with the wrong combination of colors, the unusual size of spoon, and the absence of Army soldiers’ shadows.

‍

Additionally it is noticed that the flag on Indian soldiers’ shoulder appears wrong and it is not the traditional tricolor pattern. Another anomaly, soldiers with three arms, strengtheness the idea of the AI generated image. 

Furthermore, we used the HIVE AI image detection tool and it was found that each photo was generated using an Artificial Intelligence algorithm.

We also checked with another AI Image detection tool named Isitai, it was also found to be AI-generated.

‍

‍

After thorough analysis, it was found that the claim made in each of the viral posts is misleading and fake, the recent viral images of Indian Army soldiers eating food on the border  in the extremely  hot afternoon of Badmer were generated using the AI Image creation tool.

Conclusion:

In conclusion, the analysis of the viral photographs claiming to show Indian army soldiers having their lunch in scorching heat in Barmer, Rajasthan reveals many anomalies consistent with AI-generated images. The absence of shadows, distorted hand placement, irregular showing of the Indian flag, and the presence of an extra arm on a soldier, all point to the fact that  the images are artificially created. Therefore, the claim that this image captures real-life events is debunked, emphasizing the importance of analyzing and fact-checking before sharing in the era of common widespread digital misinformation.

• Claim: The photo shows Indian army soldiers having their lunch in extreme heat near the border area in Barmer/Jaisalmer, Rajasthan.
• Claimed on: X (formerly known as Twitter), Instagram, Facebook
• Fact Check: Fake & Misleading

Introduction

Social media has emerged as a leading source of communication and information; its relevance cannot be ignored during natural disasters since it is relied upon by governments and disaster relief organisations as a tool for disseminating aid and relief-related resources and communications instantly. During disaster times, social media has emerged as a primary source for affected populations to access information on relief resources; community forums offering aid resources and official government channels for government aid have enabled efficient and timely administration of relief initiatives.

However, given the nature of social media, misinformation risks during natural disasters has also emerged as a primary concern that severely hampers aid administration during natural disasters. The disaster-disinformation network offers some sensationalised influential campaigns against communities at their most vulnerable. Victims who seek reliable resources during natural calamities often reach out to inhospitable campaigns and may experience delayed or lack of access to necessary healthcare, significantly impacting their recovery and survival. This delay can lead to worsening medical conditions and an increased death toll among those affected by the disaster. Victims may lack clear information on the appropriate agencies to seek assistance from, causing confusion and delays in receiving help. 

Misinformation Threat Landscape during Natural Disaster

During the 2018 floods in Kerala, it was noted that a fake video on water leakage from the Mullaperyar Dam created panic among the citizens and negatively impacted the rescue operations. Similarly, in 2017, reports emerged claiming that Hurricane Irma had caused sharks to be displaced onto a Florida highway. Similar stories, accompanied by the same image, resurfaced following Hurricanes Harvey and Florence. The disaster-affected nation may face international criticism and fail to receive necessary support due to its perceived inability to manage the crisis effectively. This lack of confidence from the global community can further exacerbate the challenges faced by the nation, leaving it more vulnerable and isolated in its time of need.

The spread of misinformation through social media severely hinders the administration of aid and relief operations during natural disasters since it hinders first responders' efforts to counteract and reduce the spread of misinformation, rumours, and false information and declines public trust in government, media, and non-governmental organisations (NGOs), who are often the first point of contact for both victims and officials due to their familiarity with the region and the community. In Moldova, it was noted that foreign influence has exploited the ongoing drought to create divisions between the semi-autonomous regions of Transnistria and Gagauzia and the central government in Chisinau. News coverage critical of the government leverages economic and energy insecurities to incite civil unrest in this already unstable region. Additionally, First responders may struggle to locate victims and assist them to safety, complicating rescue operations. The inability to efficiently find and evacuate those in need can result in prolonged exposure to dangerous conditions and a higher risk of injury or death.

Further, international aid from other countries could be impeded, affecting the overall relief effort. Without timely and coordinated support from the global community, the disaster response may be insufficient, leaving many needs unmet. Further, misinformation also impedes military, reducing the effectiveness of rescue and relief operations. Military assistance often plays a crucial role in disaster response, and any delays can hinder efforts to provide immediate and large-scale aid.

Misinformation also creates problems of allocation of relief resources to unaffected areas which resultantly impacts aid processes for regions in actual need. Following the April 2015 earthquake in Nepal, a Facebook post claimed that 300 houses in Dhading needed aid. Shared over 1,000 times, it reached around 350,000 people within 48 hours. The originator aimed to seek help for Ward #4’s villagers via social media. Given the average Facebook user has 350 contacts, the message was widely viewed. However, the need had already been reported on quakemap.org, a crisis-mapping database managed by Kathmandu Living Labs, a week earlier. Helping Hands, a humanitarian group was notified on May 7, and by May 11, Ward #4 received essential food and shelter. The re-sharing and sensationalisation of outdated information could have wasted relief efforts since critical resources would have been redirected to a region that had already been secured.

Policy Recommendations

Perhaps the most important step in combating misinformation during natural disasters is the increasing public education and the rapid, widespread dissemination of early warnings. This was best witnessed in the November 1970 tropical cyclone in southeastern Bangladesh, combined with a high tide, struck southeastern Bangladesh, leaving more than 300,000 people dead and 1.3 million homeless. In May 1985, when a comparable cyclone and storm surge hit the same area, local dissemination of disaster warnings was much improved and the people were better prepared to respond to them. The loss of life, while still high (at about 10,000), the numbers were about 3% of that in 1970. On a similar note, when a devastating cyclone struck the same area of Bangladesh in May 1994, fewer than 1,000 people died. In India, the 1977 cyclone in Andra Pradesh killed 10,000 people, but a similar storm in the same area 13 years later killed only 910. The dramatic difference in mortalities was owed to a new early-warning system connected with radio stations to alert people in low-lying areas. 

Additionally, location-based filtering for monitoring social media during disasters is considered as another best practice to curb misinformation. However, agencies should be aware that this method may miss local information from devices without geolocation enabled. A 2012 Georgia Tech study found that less than 1.4 percent of Twitter content is geolocated. Additionally, a study by Humanity Road and Arizona State University on Hurricane Sandy data indicated a significant decline in geolocation data during weather events.

Alternatively, Publish frequent updates to promote transparency and control the message. In emergency management and disaster recovery, digital volunteers—trusted agents who provide online support—can assist overwhelmed on-site personnel by managing the vast volume of social media data. Trained digital volunteers help direct affected individuals to critical resources and disseminate reliable information. 

Enhancing the quality of communication requires double-verifying information to eliminate ambiguity and reduce the impact of misinformation, rumors, and false information must also be emphasised. This approach helps prevent alert fatigue and "cry wolf" scenarios by ensuring that only accurate, relevant information is disseminated. Prioritizing ground truth over assumptions and swiftly releasing verified information or acknowledging the situation can bolster an agency's credibility. This credibility allows the agency to collaborate effectively with truth amplifiers. Prebunking and Debunking methods are also effective way to counter misinformation and build cognitive defenses to recognise red flags. Additionally, evaluating the relevance of various social media information is crucial for maintaining clear and effective communication.

References

1. https://www.nature.com/articles/s41598-023-40399-9#:~:text=Moreover%2C%20misinformation%20can%20create%20unnecessary,impacting%20the%20rescue%20operations29. 
2. https://www.redcross.ca/blog/2023/5/why-misinformation-is-dangerous-especially-during-disasters
3. https://www.soas.ac.uk/about/blog/disinformation-during-natural-disasters-emerging-vulnerability‍
4. https://www.dhs.gov/sites/default/files/publications/SMWG_Countering-False-Info-Social-M dia-Disasters-Emergencies_Mar2018-508.pdf