AI Election Year: What You Need to Know

The Impact of Generative AI on Global Elections

Introduction

Generative AI is increasingly being used to deceive and manipulate voters worldwide. This trend is particularly concerning in the largest global election year to date. Today on WIRED Politics Lab, we discuss a new project tracking AI’s role in elections globally. Additionally, Nilesh Christopher explores the deepfake industry and its influence on Indian voters.

The Rise of AI in Politics

Global Election Landscape

This year marks the largest global election year ever, with more people voting than ever before. Over 60 countries are holding elections, including more than 50 general elections. This surge in voter participation coincides with the widespread use of social media and the internet, making it a critical time for examining the impact of generative AI.

The Role of Social Media

Social media platforms have long struggled with election-related misinformation. Now, generative AI adds a new layer of complexity. While deepfakes and video fakes are often obvious, they represent just a fraction of the potential for AI to manipulate information.

Tracking AI’s Influence

The WIRED Project

WIRED has launched a comprehensive project to track the use of AI in elections worldwide. Reporter Vittoria Elliott leads this initiative, aiming to collect and share examples throughout the year. The project starts with around 50 instances of AI use in politics, with more to be added as they are discovered.

“We’re not seeing this as a trend, this is aberrations. These are one-off instances.”

The Uncertainty of AI’s Impact

Despite assurances from industry leaders like Nick Clegg, President of Global Affairs at Meta, the true extent of AI’s influence remains unknown. The rapid advancement of this technology makes it crucial to monitor its use and potential impact on elections.

Real-World Examples

AI in Indian Elections

Generative AI is already playing a significant role in elections around the world, particularly in India. Deepfakes and other AI-generated content are being used to influence voters, raising concerns about the integrity of the electoral process.

The Broader Implications

The use of AI in politics is not limited to deepfakes. AI can manipulate existing content, such as audio, video, and photos, or create entirely new content, like speeches. This makes it challenging to verify the authenticity of information, especially with audio and text.

The Challenges Ahead

Identifying AI-Generated Content

One of the biggest challenges is identifying AI-generated content. While some organizations, like WIRED, choose to watermark AI-generated images to indicate their origin, many new tools do not. The lack of regulation and self-policing by major platforms like Meta and Google complicates the issue further.

The Need for Regulation

As AI technology continues to evolve, there is an urgent need for regulation to prevent misuse. Platforms are currently developing policies on the fly, addressing problems as they arise. However, a more proactive approach is necessary to safeguard the integrity of elections.

Conclusion

Generative AI’s role in global elections is a growing concern. As the technology advances, it becomes increasingly important to monitor and regulate its use to ensure fair and transparent elections. The WIRED project aims to shed light on this issue, providing valuable insights into the impact of AI on the democratic process.### The Role of AI in Political Campaigns

AI-Generated Content in Campaigns

AI technology is increasingly being used in political campaigns. For example, the Democratic National Committee (DNC) created an AI-generated parody of a song by Laura Trump, co-chair of the Republican National Committee (RNC). They were transparent about its AI origins. However, not all AI-generated content is openly claimed. Often, such content appears on social media platforms like Twitter, Facebook, or WhatsApp without clear attribution.

Identifying AI Sources

Sometimes, it’s possible to trace AI-generated content back to specific companies. For instance, a deepfake of former Pakistani Prime Minister Imran Khan was created using ElevenLabs, the same company behind a fake Joe Biden robocall. However, identifying the sources is not always straightforward.

Company Policies on AI Use

Major AI companies like Midjourney, OpenAI, and Google have stated they will implement guardrails to prevent the misuse of their technologies in political campaigns. For example, OpenAI’s ChatGPT has policies against generating political content. Despite these policies, enforcement is often lacking. A report from the Center for Countering Digital Hate revealed that image generators could still produce politically charged images upon request.

Case Studies

Dean Phillips’ Chatbot

Former Congressman Dean Phillips created a chatbot called Dean.bot using OpenAI’s ChatGPT. It wasn’t taken down until media scrutiny highlighted its policy violations.

Indonesian Elections

In Indonesia, an app called Pemilu was developed to assist in elections. The app’s founder claimed it used ChatGPT to craft campaign speeches in various local languages, tailoring messages to specific demographics.

Emotional Impact of AI Content

AI-generated content can evoke strong emotions, even when people know it’s fake. For example, the far-right party AFD in Germany used a potentially AI-manipulated image in an ad to stir fears about immigrant crime. Similarly, in American Samoa, a no-name candidate, Jason Palmer, used a generative AI avatar to win the Democratic primary by sending personalized messages to voters.

“People are really worried about how technology can be used to deceive.”

Conclusion

AI’s role in political campaigns is multifaceted, from creating emotional connections to spreading misinformation. While some uses, like translating speeches, may seem benign, others can be manipulative and deceptive. The challenge lies in identifying and regulating these uses to ensure fair and transparent elections.### The Future of Generative AI in Elections

Global South Trends

Localized Messaging and Deepfakes
Generative AI is making significant inroads in the Global South, particularly in countries like Indonesia, India, Bangladesh, and Pakistan. Here, AI is being used to create highly localized messages, avatars, and deepfakes tailored to specific constituencies. For example, in Bangladesh, a deepfake video falsely showed a local politician conceding an election, highlighting the potential for misuse.

“We need to be really, really careful about this. We’re really going to be relying a lot on researchers, of whom there are not enough at this point to fact-check all this stuff, and self-admission.”

Challenges in Verification
The sheer volume of AI-generated content poses a significant challenge for fact-checkers and researchers. The effort required to verify such content is immense, and there is a growing concern that the current resources may not be sufficient to keep up.

How You Can Help

Submission Form for AI Instances
If you come across any generative AI content, you can help by submitting it through a form available on our project page. Even if you’re unsure about the content, researchers are available to help verify it.

Deepfakes in Indian Elections

Evolution of AI Use

From Jokes to Serious Campaign Tools
Initially, generative AI in India was used humorously, such as cloning Prime Minister Narendra Modi’s voice to make him sing in different languages. However, it has evolved into a serious tool for personalized voter outreach, especially in regional languages.

Rise of Deepfake Service Providers
Deepfake service providers have emerged as a $60 million industry, offering services like creating digital avatars of politicians and making AI calls in their voices. For instance, in the lead-up to the elections, 15 million calls were made to voters, either canvassing votes or reaching them on specific occasions.

Personalization and Impact

Local Politicians and Personalized Outreach
Local politicians, like Shakti Singh Rathore from Rajasthan, are using AI to create personalized messages in their voices, informing voters about government schemes and canvassing for votes. This personalized approach makes voters feel recognized and heard, especially in rural areas.

“The more calls, the more outreach you do, the more validated they feel. That is the key insight which these personalized voter outreaches are built on.”

Tech Talent and Regional Focus
India’s robust local tech sector plays a crucial role in this evolution. Startups are focusing on regional languages, creating a linguistic moat rather than a technological one. Companies like iToConnect have made millions of personalized calls, even for local politicians.

Key Figures and Industry Dynamics

Divyendra Singh Jadoun: A Case Study
Divyendra Singh Jadoun, a former politician, has become a prominent figure in this industry. Without external investors, he has the freedom to work across party lines and languages, scaling his solutions effectively.

“Talk about being profiled by New York Times, Washington Post, and everyone just because of a hobby you picked up during COVID.”

Impact and Ethical Concerns
The use of generative AI in elections is both impactful and manipulative. It raises ethical concerns about the influence of AI on voter behavior and the potential for misinformation.

Conclusion

Generative AI is rapidly transforming election campaigns, especially in the Global South. While it offers new ways to engage voters, it also poses significant challenges in terms of verification and ethical use. The involvement of local tech talent and the rise of deepfake service providers highlight the growing complexity of this issue.### The Rise of Generative AI in Politics

Early Beginnings

The first image on his Instagram shows him and his friends campaigning for elections, adorned with garlands and surrounded by armed guards. This was during his student election campaign in early 2018-2019. These young individuals wanted to be taken seriously and learned voter psychology, communication, and brand-building. During the COVID lockdown, he experimented with various activities, including learning the flute in a month.

He could have become a concert flutist, but instead went down the generative AI pathway as one does.

From Art to AI

He also tried becoming ambidextrous and used open-source tools to create fun videos for newlyweds, grafting their faces onto Superman footage. This hobby eventually led him into the political generative AI space.

So many people, the world would be so different if we let them pursue art.

The Lucrative Wedding Industry

Initial Ventures

Starting with romantic and wedding memes, he tapped into the lucrative wedding industry in India. This venture eventually expanded into political AI applications.

So much money there, yes.

Expansion Plans

Divyendra Singh, the individual profiled, has no external investors but aims for significant growth. He and his partner have pitched to international politicians, especially in Canada, which has a large Indian diaspora. They aim to use generative AI for voter outreach and are also targeting the US elections.

Regulatory Challenges

Global Expansion

The global expansion of this industry is noteworthy, especially with upcoming elections worldwide. The US has started to regulate AI, particularly after incidents like the Biden robocall.

We’re around five months out from the United States election and there’s so many other elections happening around the world.

Indian Government’s Response

India has made some attempts to regulate AI, especially after a viral deepfake incident involving an actress. The government proposed licensing generative AI models but rolled back the advisory due to its impracticality.

Impact on Elections

Chaos and Misinformation

The rise of deepfakes has caused chaos in Indian elections. Fact-checkers struggle with limited access to forensic testing AI capabilities. Local startups have emerged to address this gap, but deepfakes continue to spread misinformation.

It’s a lot of chaos. It has been a lot of chaos.

The Liar’s Dividend

The concept of the “Liar’s Dividend” is particularly concerning. When everything can be faked, even real content is questioned. This has given politicians leverage to dismiss genuine recordings as deepfakes.

That’s one of the most worrying things that we have witnessed, the erosion of trust in any material we see online, be it audio or video.

Lessons for the Future

Insights for the US

As the US heads into its election season, there are lessons to be learned from India’s experience with generative AI. The misuse of deepfakes and the resulting mistrust in media are significant challenges that need addressing.

Calling even a simple splice of a video as a deepfake causes this problem of Liar’s Dividend, where politicians can simply get away with disclaiming deepfake on any audio or video.