The AI Deception: Unveiling the Impact of Generative AI and Deepfakes on Global Elections

To go back to this article, navigate to My Profile and then click on View saved stories.

By Leah Feiger

Introducing the AI-Powered Election Season

This year marks the most extensive worldwide election period to date, with generative AI playing a significant role in deceiving and influencing voters globally. Could this escalating phenomenon make a difference? In this episode of WIRED Politics Lab, we discuss a fresh digital initiative aimed at monitoring AI's involvement in global electoral processes. Additionally, Nilesh Christopher explores the booming market of deepfakes and the strategies politicians employ to overwhelm voters in India.

Leah Feiger can be found on Twitter as @LeahFeiger, while Vittoria Elliot goes by @telliotter. You can reach out to us via email at politicslab@WIRED.com. Don't forget to sign up for the WIRED Politics Lab newsletter by following this link.

Highlighted this week: AI Global Elections Initiative

"Deepfakes Flood Indian Electorate, Garnering Approval from Political Aspirants," by Nilesh Christopher.

"US Tech Giants Continue to Host a Right-Wing Indian Outlet Spreading Racist Theories," by Vittoria Elliot and David Gilbert.

Listening Guide

To tune into this week's podcast episode, simply use the audio player featured on this site. However, for those interested in automatically receiving each episode, here's your guide to subscribing at no cost:

For iPhone or iPad users, launch the Podcasts application, or simply click on the provided link. Alternatively, you can install applications such as Overcast or Pocket Casts, and look up WIRED Politics Lab. Additionally, we are available on Spotify.

Transcript Notice: Please be aware this transcript was generated automatically and may include inaccuracies.

Leah Feiger: Hello and welcome to the WIRED Politics Lab, your go-to podcast for exploring the intersection of technology and politics. My name is Leah Feiger, and I serve as the chief politics editor here at WIRED. We're witnessing some bizarre developments in the political arena globally. Imagine if Eminem threw his weight behind a political faction in South Africa, or long-deceased leaders in India and Indonesia suddenly appeared, endorsing current electoral candidates. Picture, if you will, Joe Biden urging New Hampshire voters to stay indoors during the primaries. Of course, none of these events are real. Eminem hasn't publicly involved himself in South African political matters, at least to our knowledge. However, the power of generative AI means that creating convincing fabrications of public figures saying or doing things they never actually did is now within reach, including making false statements about elections or political campaigns. In this episode, we're diving into the ways AI is currently being utilized to influence political outcomes this year. WIRED is embarking on an extensive initiative to monitor AI's role in elections globally, gathering and disseminating instances throughout the year. Leading this ambitious effort is reporter Vittoria Elliott, who joins us today to share insights into the world of generative AI. Tori, it's reassuring to see you in person and know you're truly here.

Vittoria Elliott: I appreciate your kindness. Truth be told, I exist as a trio of raccoons stacked inside a long coat. It's the raccoon at the center that manipulates my hands.

I'm sorry,

Authored by Reece Rogers

Authored by Will

Authored by Kim Zetter

Leah Feiger: Tori, what prompted you to propose this project to me initially?

Vittoria Elliott: Absolutely. This year's elections are monumental. Globally, we're witnessing an unprecedented number of people casting their votes. This includes a variety of elections, from local to parliamentary, and crucially, numerous general elections. Over 60 countries are participating in the electoral process, with general elections occurring in more than 50 of these. This entails a wide array of voting activities and, in many instances, the selection of national leaders.

Leah Feiger: This is the highest number of elections to date.

Vittoria Elliott highlighted that this election has seen the highest voter turnout and is the largest one since the advent of social media, likely marking the greatest online participation during an electoral period to date.

Leah Feiger: Exactly.

Vittoria Elliott: The challenge is enormous. For years, social media platforms have grappled with the complex problems posed by elections and political content, especially concerning misinformation and disinformation. Currently, we're encountering an additional complication: generative AI. The situation is particularly difficult at the moment, as deepfakes and fake videos can be quite apparent. However, these examples only represent the most visible aspects of a much larger issue.

Leah Feiger: So, to what extent do we expect generative AI to play a role in electoral processes? Or is it premature to make any predictions?

Vittoria Elliott highlighted an interesting point, mentioning that Nick Clegg, Meta's President of Global Affairs (the company behind Facebook and Instagram), recently downplayed the significance of generative AI in politics. Clegg described these occurrences as isolated incidents rather than a trend. However, there's a clear acknowledgment of the unknowns surrounding the technology's application in politics. To address this, a project was initiated to catalog examples of generative AI's political use worldwide, starting with a collection of 50 instances. Elliott points out that this is just the beginning, with more data being compiled. The initiative stems from the uncertainty of how generative AI might influence future elections globally, emphasizing the importance of understanding its potential impact. Given the rapid evolution of this technology and its observable effects in recent elections, there's a push to grasp the changing tactics and their implications by year's end.

Leah Feiger: It's evident that in the recent and ongoing elections globally, generative AI is significantly influencing outcomes, notably in countries such as India. We'll delve deeper into this topic in the upcoming segment, so make sure to keep watching. It's undeniable that this trend is unfolding.

Vittoria Elliott expressed concern, emphasizing that the true alarm lies in our ignorance of the full scope. Often, when folks consider generative AI, their minds go to ChatGPT producing nonsensical responses or the notably poor search results emerging from Google lately.

Authored by Tess

Authored by Reece Rogers

Authored by Will

Authored by Kim Zetter

Leah Feiger: Advising you to add adhesive to your pizza.

Vittoria Elliott expressed, "I actually came across that one today and thought to myself, 'My friends.'"

Leah Feiger: Truly excellent. Google offered some really solid recipe suggestions. But seriously, avoid doing that, everyone.

Vittoria Elliott remarked that the common perception of generative AI, such as ChatGPT or the creation of deepfakes where fabricated videos make it seem as though politicians are uttering statements they never actually made, resembles the naivety of a child eating glue. These examples, though significant, only scratch the surface of what generative AI is capable of. It has the power to alter existing media, including audio, visuals, and photographs, as well as to create entirely new content from scratch, like composing a speech upon request—a capability that has already been demonstrated. The challenge we face today is that it's significantly simpler to discern the authenticity of a video, which usually involves complex, verifiable events, compared to other forms of media.

Leah Feiger: Agreed.

Vittoria Elliott highlights the challenges in detecting deepfakes when they are in audio or textual format, noting their prevalence in countries like India and Indonesia, and significantly in Pakistan and Bangladesh. She points out that while visual deepfakes are somewhat easier for individuals to identify, these instances could merely represent a small fraction of what's actually happening.

Leah Feiger: Who's really behind the production and online posting of this material?

Vittoria Elliott: That's an excellent inquiry. Often, it's unclear to us. Within the realm of artificial intelligence, there exists a concept known as watermarking. This allows for the marking of creations to indicate that they have been produced by generative AI.

Leah Feiger: Exactly. For the recent project we released, we made sure to place a prominent watermark on all the AI-generated images we used, clearly stating, “This content is generated. It's not real.”

Vittoria Elliott: Precisely. However, it's a decision we've made voluntarily. It's an act of kindness from us since we aim to prevent its misuse by others.

Leah Feiger: We excel in what we do.

Vittoria Elliott comments, "Indeed, we are outstanding individuals. However, numerous emerging technologies lack oversight. There's a scarcity of regulations in this area. Even the larger platforms are monitoring themselves. It's like they are constructing the path while simultaneously walking on it. Companies such as Meta and Google are implementing rules as they begin to recognize potential problems that could have serious consequences. Often, it's not clear who is behind these creations. In some cases, creators will acknowledge their work, especially if it's produced for a specific campaign. For example, the Democratic National Committee openly created an AI spoof involving Laura Trump, who serves as the co-chair for the Republican National Committee, mocking a song she released. They were upfront, stating it was generated by AI. We're aware in instances like these, they'll take ownership. However, there are many cases where the origin of content we encounter online is unclaimed. It may circulate on social media platforms like Twitter, Facebook, or WhatsApp, and it's challenging to verify the true creator. We only know which account has disseminated it."

Authored by Tess

Authored by Reece Rogers

Authored by Will

Authored by Kim Zetter

Leah Feiger: At times, it's possible to trace it directly to certain businesses—

Vittoria Elliott: Affirmative.

Leah Feiger: … involved in creating the generative AI technology.

Vittoria Elliott: Absolutely. For example, a deepfake of Imran Khan, the ex-prime minister of Pakistan who has been imprisoned on corruption allegations, was created. His political party was barred from participating in the general elections earlier this year. Despite this, he managed to deliver campaign speeches through the use of generative AI technology.

Leah Feiger: Untamed.

Vittoria Elliott mentioned that for this purpose, ElevenLabs was utilized, which is the identical firm that had been employed for the counterfeit Joe Biden automated call earlier in the year. In some instances, the companies engaged are known, but often, their identities remain undisclosed.

Leah Feiger: What strategies have these corporations announced for dealing with elections this year?

Vittoria Elliott noted that reputable organizations such as Midjourney, ChatGPT by OpenAI, Google, and others have committed to implementing safeguards to prevent the creation of political imagery. Specifically, ChatGPT, which focuses on text, has made it clear that its platform should not be used to produce political content for campaigns or to operate chatbots using its framework. However, their enforcement efforts appear to be lacking. A study conducted by the Center for Countering Digital Hate, which we discussed in March, revealed that when these image-generating platforms were asked to produce images of political figures like Trump or Biden engaging in various activities, they often complied. Regarding ChatGPT, there was a mention of Dean Phillips, a congressman who had briefly entered the presidential race.

Leah Feiger: Previously a presidential candidate, Representative Dean Phillips.

Vittoria Elliott: …created a chatbot named Dean.bot using OpenAI's ChatGPT framework, and it remained active until the media pointed out, “Doesn’t this violate your guidelines?”

Leah Feiger recalls vividly. Another detail that stands out to her from that time is Dean Phillips' significant support from Silicon Valley investors. The situation seems somewhat unclear. It's as though there's a mixed message of disapproval for using Dean.bot, yet a continued affection and backing. This creates a peculiar push-pull dynamic. For instance, Dean's comments about regulating generative AI, which Sam Altman showed interest in, highlight this contrast.

Vittoria Elliott mentioned that the scenario was specific to the United States. For example, in Indonesia, there's a firm that created an application known as Pemilu tailored for the Indonesian electoral process. The individual who initiated this project asserted they had developed an enhancement for ChatGPT, enabling it to craft political speeches in various indigenous dialects.

Leah Feiger: Astonishing.

Vittoria Elliott explained that this involved gathering data to customize messages for specific groups, such as the youth, women, and others.

Leah Feiger: Indeed, it's quite impactful in a nation where a multitude of languages are spoken.

Vittoria Elliott mentioned that the distribution spans numerous islands, each with distinct requirements.

I'm sorry,

Since you haven't provided any

Authored by Will

Authored by Kim Zetter

Leah Feiger: Without a doubt.

Vittoria Elliott reported that a man openly admitted to Reuters, saying, "Yes, we developed it using ChatGPT as a base." She questioned whether OpenAI would have ever identified how their technology was being utilized.

Leah Feiger: Correct.

Vittoria Elliott: It remains uncertain.

Leah Feiger: Indeed, as you mentioned, identifying these items is quite challenging.

Vittoria Elliott mentioned, "Indeed, the methods we employ are still in their early stages, and they're often optional."

Leah Feiger: Indeed, they depend on individuals voluntarily coming forward and those who are cooperative. Individuals who are ready to admit, "Yes, I've utilized this." Or question, "Is it permissible for me to use it in this manner?" Fascinating. There's quite a variety presented here, with some scenarios seeming quite alarming, while others appear less so. However, converting a speech into another language seems relatively benign to me. It essentially involves providing information to voters.

Vittoria Elliott: Absolutely. Much of the discussion surrounding generative AI revolves around concerns of misinformation and disinformation. There's a significant amount of worry regarding the potential for technology to mislead.

Leah Feiger: Definitely. That's a possibility. As we mentioned earlier, there might be some South African voters who were extremely excited to learn that Eminem was backing their preferred political figure.

Vittoria Elliott mentioned that besides its other strengths, it can serve as a tool for satire. In fact, she pointed out that in India, its origins were rooted in satire, a topic she plans to delve into more in the second segment. It has the capability to mobilize individuals. However, she highlighted a dilemma: even when individuals are aware that certain content is fabricated, it can still evoke strong emotions. This is particularly true when one encounters visuals or statements that resonate on a personal level, regardless of the awareness of their authenticity.

Leah Feiger: Absolutely.

Vittoria Elliott: That evokes a strong emotional response.

Leah Feiger: Can you provide a specific example?

Vittoria Elliott recently covered a story focusing on the AFD, a far-right political group in Germany. This group has been flagged by a German court as showing signs of anti-democratic and extremist tendencies. In their recent advertising campaign on social media platforms like Facebook and Instagram, the AFD used an image depicting a white female with facial injuries. According to research, this image is likely to have been altered using advanced AI technology. Accompanying the image was a message claiming an increase in crime due to immigrants. This kind of imagery and messaging could resonate with individuals already harboring fears or biases, reinforcing their concerns about crime related to immigrants, especially when such claims are echoed by political figures.

Leah Feiger: It would frighten you.

Vittoria Elliott: This experience can evoke strong emotions, as it might resonate with your existing feelings or convictions.

Leah Feiger: Absolutely. Despite the fact that deceased political figures are reappearing in India and Indonesia to address their voters, the public is aware that these politicians have passed away. It's quite heartening to see them return and endorse a candidate by saying, “Please support XYZ candidate.” There's an emotional connection and appeal to it, even though everyone understands it's not real.

Authored by Tess

Authored by Reece Rogers

Authored by Will

Authored by Kim Zetter

Vittoria Elliott: Indeed. A prime illustration can be found in American Samoa, where an obscure contender, Jason Palmer, an American entrepreneur with no prior political experience, emerged victorious in the Democratic primary. This success came about because he utilized a digital representation of himself to interact with and respond to inquiries from the electorate.

Leah Feiger: Regarding the topic of generative artificial intelligence?

Vittoria Elliott: Discussing the potential of generative artificial intelligence.

Leah Feiger: Astonishing.

Vittoria Elliott reported that he employed generative artificial intelligence to dispatch customized messages via text and email to individuals. His success in the primary was notable. She speculated that recipients viewing his digital avatar might not have believed it accurately represented him, especially in responding to their inquiries instantly. Elliott believes that people are aware that communications received from an official political campaign are indeed part of the campaign's strategy.

Leah Feiger: Agreed.

Vittoria Elliott: It's designed to influence their decision.

Leah Feiger: Absolutely.

Vittoria Elliott: However, experiencing that kind of bond still evokes a strong emotional response.

Leah Feiger: Indeed, predicting what lies several months down the line is challenging. However, based on your recent compilation of data and examples, how do you envision the trajectory of generative AI in the context of future elections? What expectations or forecasts do you hold for its role and impact?

Vittoria Elliott emphasizes the significant trend of using avatars and deepfakes for targeted political messaging in countries of the Global South, such as Indonesia, India, Bangladesh, and Pakistan. She points out that these digital manipulations are particularly effective and dangerous because they are crafted to appeal to specific local groups. A notable example Elliott mentions is from Bangladesh, where a deepfake video falsely showed a female politician conceding an election. This highlights the critical challenge ahead: the need for vigilant fact-checking and research to combat misinformation. However, Elliott expresses concern over the shortage of researchers available to tackle the massive amount of content that needs verification. Despite the efforts to catalog and confirm instances of misinformation, she fears that the sheer volume of work required may overwhelm the capacity to effectively counteract these deceptive practices.

Leah Feiger: Ending on such a positive note, Tori, how can individuals contribute to our initiative?

Vittoria Elliott: Have you come across any examples of generative AI in your daily life? We've set up a submission form in the show notes and on our project's webpage for you to share these instances with us. Our goal is to monitor these occurrences throughout the year. Feel free to submit anything you find puzzling or uncertain about, as we've teamed up with several researchers who are ready to assist us in verifying these submissions. Whether it's an unusual video, an odd voice recording, or a strange message that has you questioning its authenticity, don't hesitate to forward it to us for review.

I'm sorry,

—

Authored by Reece

Authored by Will

Authored by Kim Zetter

Leah Feiger: Appreciate it, Tori. We're going to take a short break and return shortly.

Unfortunately, without the

Leah Feiger: Greetings once again to WIRED Politics Lab. Tori remains here with us in the studio. Today, we have the pleasure of having Nilesh Christopher with us, a contributor to WIRED who has recently published an article focusing on the use of deepfakes in India's electoral process. Nilesh, we're truly grateful for your presence today.

Nilesh Christopher: I appreciate the opportunity to join you.

Leah Feiger: Could you share your experiences? What observations have you made? In what ways are politicians in India and their campaign strategies incorporating generative artificial intelligence?

Nilesh Christopher observed with great interest how the narrative around generative AI has transformed. Initially, it was employed for humorous purposes. For instance, Prime Minister Narendra Modi's voice was replicated to humorously make him sing in languages he's not fluent in.

Leah Feiger: Agreed.

Nilesh Christopher discussed how voter engagement is being customized through various regional dialects, a trend observed over the last half-year. The focus of our investigation was on the emergence of companies specializing in the authorized creation of political deepfakes, which provide tailored communication services. This sector has burgeoned into a lucrative field, now valued at $60 million, encompassing the generation of digital representations of political figures for promotional purposes to the utilization of artificial intelligence for making phone calls in the voices of these politicians. Notably, in the buildup to the elections this April, a staggering 15 million calls were made directly to voters, serving either to solicit their votes or to contact them on special occasions.

Leah Feiger: That's incredible. It's occurring among politicians across the board, isn't it? They're engaging firms to supply them with deepfake representations of themselves.

Nilesh Christopher: Indeed, Leah, what caught me off guard was a particular case involving a local leader, Shakti Singh Rathore, who wasn't faring well in the polls. He's from Rajasthan, a state in the northern part of India, and a member of the Bharatiya Janata Party, which is led by Modi and is known for its Hindu nationalist stance. Rathore aimed to amplify the accomplishments of Modi's tenure over the last decade, so he reached out to his friend, 31-year-old Divyendra Singh Jadoun from Ajmer, a small locality. Their plan was innovative: Rathore intended to replicate his persona, creating personalized voice messages to inform the voters in his district about the initiatives undertaken by Modi, like Digital India, which focuses on various development projects for the Indian populace. The goal was to enlighten the electorate about the party's achievements, hoping to secure their support in the forthcoming elections.

Leah Feiger noted the intimate nature of the campaign tactics described, highlighting the impact of personalized outreach on voters. She found it particularly compelling that in a nation as vast and populous as the world's largest democracy, amidst numerous elections and extensive campaigning periods, efforts were made to create a sense of individual recognition and connection. This was achieved through personalized calls, multimedia content, linguistic adaptations, and even music, Feiger observed.

Authored by Tess

Authored by Reece Rogers

Authored by Will

By Kim Zetter

Nilesh Christopher: Definitely. It's interesting to note that the majority of residents in major Indian cities such as Bangalore or Delhi tend to ignore or dismiss spam calls or robocalls instead of answering them.

Leah Feiger: Absolutely, I believe that's a common sentiment.

Nilesh Christopher: Definitely. What's fascinating is that residents in smaller communities, particularly in tier-two and tier-three cities, experience a feeling of acknowledgment and attention from their local representatives when they receive these calls. The greater the number of calls made, the more outreach efforts are extended, the stronger their sense of validation becomes. This fundamental understanding underpins the strategy behind these customized voter engagement campaigns. We had a conversation with someone who manages a firm called iToConnect, which conducted close to 25 million personalized calls…

Leah Feiger: Astonishing!

Nilesh Christopher discusses how the outreach efforts are not limited to prominent national figures such as Modi or similar politicians. Instead, local leaders are actively engaging with the electorate within their own regions, seeking and soliciting votes from their constituents.

Vittoria Elliott: Nilesh, given India's strong domestic technology industry, could you delve into its impact? Specifically, how does the abundance of tech expertise in India contribute to the development in this area? Please provide insight into the participants entering the deepfake sector, which appears to be experiencing significant demand.

Nilesh Christopher: Observing this development has been intriguing, Tori, for a couple of reasons. Firstly, the diversity of languages in India, with 22 recognized officially and several thousand more, has led to the emergence of localized platforms. For example, there's been a startup focused on serving the Tamil-speaking community by compiling clips of Indian politicians who speak Tamil for their campaigns. Similarly, there are services dedicated to Hindi and Telugu speakers.

Leah Feiger: That's quite fascinating.

Nilesh Christopher mentioned that these companies lack a technological advantage, so they're creating a barrier through language proficiency instead. He observed that entities managing to execute large-scale operations, involving 50 to 60 million customized calls using various political figures' voices, prefer to remain anonymous. This secrecy is partly because they are backed by significant financial investments from outside contributors.

Vittoria Elliott: That's intriguing.

Nilesh Christopher introduces us to Divyendra Singh Jadoun, a figure who now epitomizes the sector. Originally a politician, Jadoun has transitioned to serving his colleagues in the industry without the backing of external financiers. This independence grants him the flexibility to collaborate with a diverse range of politicians, transcending party affiliations and linguistic barriers, thereby expanding his initiatives.

Leah Feiger: You describe him in a way that makes it seem like his involvement was somewhat accidental. Initially a politician, he found himself at home during the Covid pandemic, exploring various hobbies and searching for something to pass the time. It's fascinating and somewhat unexpected that he would delve into creating generative AI for political campaigns, a field both influential and, to be frank, manipulative.

Authored by Tess

Authored by Reece Rogers

Authored by Will

Authored by Kim Zetter

Nilesh Christopher: Absolutely. It's quite something to be featured by major publications like The New York Times and The Washington Post, all because of a pastime you started during the Covid pandemic.

Leah Feiger: Indeed, but it's a hobby that greatly impacts. Could you share more details about him? How did it feel to conduct an interview with someone at the forefront of this movement?

Nilesh Christopher has quite the intriguing history. His Instagram debut features a photo with him and his buddies, all wearing garlands, during an election campaign. In the snapshot, he's flanked by four armed security personnel, each wielding rifles and shotguns. Essentially—

Leah Feiger: Earnest.

Nilesh Christopher: Absolutely. Their goal was to come across as influential figures. This occurred around the time he was participating in student elections between 2018 and 2019. Picture them as young individuals striving for recognition and respect from everyone. That marked the beginning of his involvement in student politics. It was then that he began to understand voter behavior, effective communication strategies, and the importance of creating a personal brand. During the pandemic, he found himself confined to his home. He was eager to try new things. He even performed a flute piece for us during our visit, having picked up the instrument in just one month.

Leah Feiger: Instead of pursuing a career as a professional flute player, he chose to explore the field of generative artificial intelligence.

Vittoria Elliott believes that the global landscape could transform significantly if individuals were given the freedom to engage in artistic endeavors.

Nilesh Christopher was intrigued when he explored the concept of ambidexterity, prompting him to practice writing with both hands. His focus then shifted towards leveraging open-source tools accessible to everyone. With these tools, he embarked on creating entertaining content for newlyweds, crafting videos where they appeared as Superman. He managed this by collecting requests through Instagram, where he would superimpose the couple's faces onto footage of Superman, before sharing these personalized videos with their significant others. This unique venture marked the beginning of his journey, which has since evolved into his current involvement with political generative AI projects.

Vittoria Elliott remarked that given the enormity of the wedding sector in India, it makes perfect sense for him to begin by creating memes related to romance and weddings.

Nilesh Christopher: Indeed, a substantial amount of money is present.

Vittoria Elliott expressed curiosity, particularly about the influx of capital into the sector. She was eager to understand the sources of this investment. Given that technology firms anticipate expansion, she queried about the future prospects.

Nilesh Christopher discussed an individual named Divyendra Singh, who, despite lacking funding from external investors, possesses a significant desire for expansion. Singh and his business associate have actively sought the attention of various global politicians, with a particular focus on Canada due to its substantial Indian community. They approached Mr. Poilievre, a Canadian politician known to have support from the Indian electorate, with a proposal to use generative AI to dub his videos for better engagement with voters. Additionally, they have set their sights on the American electoral market, aiming to offer services that could revolutionize political campaigning. Singh's vision includes offering U.S. politicians the ability to create digital versions of themselves to send tailored video messages to their constituents. Christopher inquired about their plans for audio cloning technology in voice calls, especially since it's prohibited in the U.S. Despite the popularity of personalized audio messages from politicians in India, Singh’s team has decided not to market this particular service in the U.S. due to the legal restrictions.

Unfortunately, you didn

Certainly, here is a rewritten

Authored by Will

Authored by Kim Zetter

Leah Feiger: It's truly remarkable how this sector is growing internationally; it's definitely something we should monitor closely. With the United States election just about five months away, and numerous other elections taking place globally, it's understandable that following the conclusion of the election in India, their focus would shift towards involvement in additional campaigns.

Vittoria Elliott: Nilesh, you previously noted that in the US, there's been a noticeable concern, particularly following the Biden robocall incident, regarding the impact of this technology on our electoral processes. Similarly, India has been taking steps to impose regulations on AI as elections approach. I'm intrigued to know your thoughts on how we can reconcile the rapid growth of the sector with the government's attempts, albeit somewhat superficial, to establish certain controls over it?

Nilesh Christopher remarked on the situation, noting that while there's been plenty of talk, concrete actions from the Indian government have been lacking. He pointed out that typically, it takes a particularly shocking event for the government to react. A vivid example of this occurred last December, when a deepfake video featuring the face of an Indian actress superimposed on another woman wearing a revealing dress spread rapidly online, causing widespread outrage and concern over the technology's misuse. Following the incident, the government hastily issued a directive requiring all AI-generated models, whether developed by local startups or international entities, to obtain a government license—a mandate that was met with resistance and viewed as impractical given the government's processing capabilities. Fortunately, this directive was later withdrawn. Despite these reactionary measures, there has yet to be a significant, effective response to combat the problematic rise of deepfake technology.

Leah Feiger: How do you believe the prevalence of deepfakes is impacting electoral behavior in India, particularly in terms of voting patterns? How is this shaping the evolving landscape of elections?

Nilesh Christopher described the situation as highly chaotic, emphasizing the turmoil experienced by Indian fact-checkers due to their limited access to AI tools for forensic analysis. These fact-checkers, in the lead-up to the election, had no choice but to seek expertise from consultants or academics affiliated with universities in the United States to scrutinize dubious videos and audios. As the election approached, they found themselves without any formidable means of verification. However, the last few months have seen some improvement in what experts term the 'detection equity gap,' with the rise of local startups in India developing tools to identify AI-manipulated content. Despite these advancements, the recent discovery of three manipulated pieces of content, including audio and video deepfakes, has raised concerns. Some of these manipulated contents falsely declare victory for local parties in certain areas. While these deepfakes may not be of the highest quality, their audio manipulations are sufficiently believable to potentially spread widely.

Authored by Tess

Authored by Reece Rogers

I'm sorry,

Authored by Kim Zetter

Vittoria Elliott: Nilesh, I'd like to zoom out for a moment and discuss a concept known as the “liar's dividend.” It's this notion that in an era where anything can be fabricated, including videos and audio, authentic materials can also be dismissed as AI-generated. You've explored this topic through your work, and I'm eager to hear about your insights, particularly how this phenomenon is manifesting in India.

Nilesh Christopher expressed concern over the growing skepticism towards the authenticity of online content, whether it's audio or visual. The automatic assumption that content might be fabricated has empowered politicians who are caught making outrageous statements. During the recent elections, an incident involving a politician from the Bharatiya Janata Party (BJP), the party Prime Minister Modi is a part of, highlighted this issue. A BJP representative made a controversial comment suggesting Modi remains unmarried to sell the nation. Although it appeared to be a harmless remark initially, the politician sought to retract it after it gained traction online. Some social media accounts even speculated that the video was manipulated using deepfake technology.

Leah Feiger: Amazing.

Nilesh Christopher: Subsequently, verifiers pursued the journalist who captured the audio clip, to confirm its authenticity, ultimately establishing that it was indeed genuine.

Leah Feiger: It's incredibly surreal. It truly feels like questioning the nature of reality itself, as if we're living in the Matrix, with all online trust completely dissolved.

Nilesh Christopher: On one hand, as the situation progresses, my grasp is that the initial adoption of generative AI was primarily in creating memes and satirical content. There were instances where videos depicted the prime minister singing in various languages. This widespread exposure helped in normalizing the use of AI, paving the way for what we now recognize as the issue of the liar's dividend.

Leah Feiger: As we approach the task of reporting on the U.S. elections and related topics, I'm curious about the insights you can share. Although you're currently engaged in this work, what wisdom can you impart? What insights can the United States gain from India's experience in these matters?

Nilesh Christopher remarked, "That's a captivating inquiry, and I've been pondering it deeply. Firstly, while there's been a surge in highly realistic malicious deepfake audio and videos, we've also observed scenarios where any basic modification of audio or video is labeled as a deepfake. This tendency to label even straightforward edits as deepfakes leads to the issue known as the liar's dividend, enabling politicians to easily dismiss any audio or video by attributing it to deepfakes. Another aspect is the importance of media literacy. As journalists and fact-checkers, it's crucial for us to distinguish between what truly constitutes a deepfake and what is merely a basic edit."

Leah Feiger: Absolutely.

Nilesh Christopher emphasizes two main points: Firstly, the importance of creating an impact. Secondly, he highlights the necessity for newsrooms and journalists to have access to forensic analysis tools. This will enable them to effectively counteract the influx of both low-quality fabrications and sophisticated deepfake technologies as they emerge.

Authored by Tess

Authored by Reece Rogers

Authored by Will

Authored by Kim Zetter

Leah Feiger: Approaching us, it feels quite foreboding.

Vittoria Elliott: Bark.

Leah Feiger: Absolutely, and it's likely very accurate. Nilesh, we're grateful for your presence today. To our listeners, if you come across examples of generative AI being used in this election year, no matter where you are globally, please let us know via the submission link in our show notes. Your insights are invaluable to us. Following our break, Tori and Nilesh will share the conspiracy theories that caught their attention the most this week.

It appears that there

Leah Feiger: Glad to have you back on Conspiracy of the Week. This segment allows our guests to share the most interesting theory they've stumbled upon over the week, and then I'll choose the one that stands out the most. Tori, what have you discovered?

Vittoria Elliott: Recently, David, alongside myself, conducted an investigation into OpIndia, a website known for its right-wing stance and alignment with the BJP, notorious for disseminating Islamophobic narratives. Among these is the concept of Love Jihad, a theory suggesting that Muslim men are deliberately marrying or abducting Hindu women to alter the population's religious demographics. Our extensive coverage seemingly pressured them to adopt a subscription-based model.

Leah Feiger: Check out the story linked in our show notes, folks, it's worth your time.

Vittoria Elliott: However, what captivates me is uncovering a regional twist on a global conspiracy theory. During our investigation for the article, we stumbled upon a piece from OpIndia that had been making the rounds, which centered on George Soros, a figure who consistently finds himself in the crosshairs of right-wing groups globally—

Leah Feiger: Honestly, it's quite remarkable how it has gained international recognition.

Vittoria Elliott: Exactly, he's omnipresent. In truth, his main goal isn't what the US right-wing suggests, which is him being a member of an international conspiracy. They frequently highlight his Jewish heritage and immigrant status. However, in India, the emphasis is on his being a Caucasian male.

Leah Feiger: Absolutely.

Vittoria Elliott stated that he is supporting projects that are against India and Hinduism.

Leah Feiger: Incredible.

Vittoria Elliott: He is fundamentally committed to weakening India and its Hindu culture. It's always fascinating to observe how these international personalities adopt a local perspective in terms of their involvement in the conspiracy here.

Leah Feiger: Regarding our current circumstances.

Vittoria Elliott: Absolutely.

Leah Feiger: Discussing the worldwide spread of conspiracy theories is a topic I could endlessly delve into. Poor George is completely clueless about it.

Vittoria Elliott mentioned that his sole interest lies in supporting various democracy projects.

Leah Feiger: That's incredible. OK, I like that one. That caught me off guard. Fantastic. Nilesh, what about you? What's your conspiracy theory for this week?

Nilesh Christopher: My intention was to discuss the topic of Love Jihad, as it has captured widespread media attention.

Leah Feiger: Absolutely.

Nilesh Christopher discussed a particular concept that caught his attention, known as Vote Jihad, which was even mentioned by Prime Minister Narendra Modi during a recent political rally. This mention has given some legitimacy to a prevalent conspiracy theory suggesting that Muslim voters will unite in large numbers to vote in a way that could potentially threaten the Hindu character of India and jeopardize Hindu rights over time. This theory is tied to concerns about India's demographics, where Muslims currently make up less than 15 percent of the population. According to Tori, this notion is not unique to India but is a global concern among many in the Hindu right-wing community. They often express their worries through WhatsApp messages, warning fellow Hindus about the possibility of Muslims voting en masse to defeat Hindu candidates or to dilute Hindu rights.

Authored by Tess

Authored by Reece Rogers

Authored by Will

Authored by Kim Zetter

Leah Feiger: Once more, we witness the spread of globalization through the lens of the Great Replacement conspiracy theory. This intensely racist belief that "Immigrants are entering the country to cast votes for Democrats" is being spread by the GOP, and similar sentiments are found in India. It's quite logical, actually. It's heartening to see nations across the globe bonding over their fascination with conspiracy theories. What an intriguing note to conclude on. Oh, dear. Apologies, everyone. This time, Tori, I'm passing the baton to Nilesh.

Vittoria Elliott: That seems reasonable.

Leah Feiger: That's quite an intriguing theory.

Vittoria Elliott: They're hosting him.

Leah Feiger: And he's the one joining us. You can see, guests receive the best treatment here at WIRED Politics Lab.

Nilesh Christopher: Appreciate it, everyone.

Leah Feiger: It's been a pleasure having you with us for this insightful discussion. We appreciate you tuning into WIRED Politics Lab. If you enjoyed our conversation today, we encourage you to subscribe and leave a rating on whichever podcast platform you use. Don't forget to check out our weekly newsletter penned by Makena Kelly. You can find the newsletter link and references to the WIRED articles we talked about today in the episode's description. For any inquiries, feedback, or ideas for future episodes, feel free to reach out to us at politicslab@wired.com. We're looking forward to hearing from you. The production team for WIRED Politics Lab includes Jake Harper as producer, with Jake Lummus handling studio engineering. Amar Lal was responsible for mixing today's episode. Our executive production is led by Stephanie Kariuki, with Jordan Bell overseeing development. Chris Bannon serves as the global head of audio at Condé Nast. I'm Leah Feiger, your host, and we're eager to bring you another episode next week.

Image credit: Getty Images

Recommended for You …

Delivered to your email: Will Knight's Fast Forward delves into the progress of artificial intelligence

Millions of deepfake videos are flooding Indian electorates

Incarcerated individuals purchased tablets, only to discover an unfulfilled commitment

The single factor impeding the progress of the heat pump

Eternal Sunshine: Discover the Top Sunglasses for Any Adventure

Vittoria Elliott

Anna Merlan

Rogers Reece

Gilbert David

Dan Gearino

William Turton

Name:

Eric Ravenscraft

Reece Rogers

Additional Stories from WIRED

Evaluations and Manuals

Copyright © 2024 Condé Nast. All rights reserved. When you buy products via our website, WIRED may receive a share of the revenue, a result of our Affiliate Agreements with store partners. Replicating, sharing, broadcasting, storing, or utilizing the content from this website in any form is strictly prohibited without the explicit consent of Condé Nast. Advertisement Choices

Choose a global website