How deepfakes could impact Indian elections

Generative AI has advanced to a point where it can make us doubt the veracity of what we see on a daily basis. In electoral politics, audio and video cloning can be put to dangerous use, spreading misinformation in an all-new effective way. Just clone the voice of any political leader, superimpose the face onto an existing video clip, and share!

Let’s have a look at this video.

This video is fake, and was created by India Today’s OSINT team using a free AI tool to demonstrate its possible misuse. This ‘cheapfake’ video could easily be upgraded into a ‘deepfake’ if one uses some advanced tools. However, it precisely highlights the dangers of this rapidly improving generative AI in the upcoming elections in India, where more than half of Internet users, as per a Kantar-Google report, consume news online.

During this investigation, we came across service providers promising higher reach and impact for political parties in elections through AI solutions. One such platform is The Indian Deepfaker, which offers personalised political ad campaigns in packages costing $199 to $599. In a video ad, it said it can help parties and leaders connect with workers and voters “deeply with personal touch”.

“Using groundbreaking AI technology, lip-synching, and voice cloning, we make it possible to send audio and video messages to large audiences with their names – giving them a personalised feeling,” the promotional video goes on to say.

Social media is already crowded with deepfake content targeting Indian politicians and political parties, with an aim of swaying voter sentiments. According to this year's State of Deepfakes report, India is the sixth most vulnerable country to deepfake.

“Politicians and celebrities are especially susceptible to this threat. Their extensive availability of photos, videos, and voice data in the public domain makes them prime targets,” said Mayank Satnalika, the engineering lead at CloudSEK, a company specializing in contextual AI for cyber threat prediction, told India Today.

PM Narendra Modi has called deepfakes “one of the biggest threats facing India’s system”.

In election-bound Madhya Pradesh, two doctored video clips of superstar Amitabh Bachchan on 'Kaun Banega Crorepati' calling Chief Minister Shivraj Singh Chouhan a liar, and portraying the Congress party’s CM hopeful Kamal Nath in a positive light.

In the US, deepfake technology is already being used to create fake films, clone a candidate’s voice, or churn out a bogus narrative to undermine the opposition’s message.

In April, Donald Trump’s Republican Party published a political ad showing a dystopian scenario. It showed President Joe Biden being re-elected, followed by explosions in Taiwan caused by a Chinese invasion, waves of migrants causing panic in the US and martial law being imposed in San Francisco.

As per a New York Times report, AI is everywhere in the ongoing elections in Argentina. The country’s social media is bombarded with AI-generated images and videos – including deepfakes, and AI illustrations decorating important streets.

“India will be no differentâ€æ Media consumption on social media contributes significantly to opinion shaping and generative AI will have a major impact on upcoming Indian elections,” said Satnalika of CloudSEK.

India first faced the threat of AI interference in elections during the Delhi Assembly polls in 2020 when users encountered videos of then state BJP chief Manoj Tiwari criticizing CM Arvind Kejriwal’s policies in different languages. Massachusetts Institute of Technology (MIT) had ascertained that the videos were created using AI. The tech of generative AI and deepfake was nascent back then, but it is easily available for any Internet user today.

“The accessibility and decreased cost of large scale content creation can be manipulated by bad actors looking to disrupt elections, facilitated by generative AI,” an IT company report notes. Many say it will further deprive regional and small parties of the level playing field in elections. Big political parties with robust funding, manpower and technology will use deepfake technology to multiply their propaganda in several languages.

A case in point is PM Modi’s AI-created videos singing popular songs in several regional languages. Experts point out that videos like these add a sense of a regional touch to voters who speak a language that the prime minister does not speak. It can also “add fake audio of trusted politicians or public figures on robocalls to make misleading messages more credible”, opines Carah Ong Whaley, academic programme officer for the University of Virginia Centre for Politics.

Digvijaya Rana, a lecturer at Jindal Global University, pointed out, “We must not forget that India has the cheapest data rates in the world, which makes sharing media online more affordable, and hence it makes us one of the most vulnerable nations.”

The perils of deepfakes have made election watchdogs worried. The UK’s National Cyber Security Centre, in an annual report, has listed the rise of AI and the evolving geopolitical landscape as “significant areas of risk to UK electoral processes”.

What is more alarming is the lack of awareness among Indian policymakers. When India Today asked Union minister and Telangana BJP chief G Kishan Reddy about the circulation of synthetic election videos, he said, “I am not aware. I need to see them.”

The paramount challenge related to deepfakes is its identification. “The quality of deepfakes is so good that even forensic labs can take days to establish the veracity of a medium,” adds cyber security expert Jiten Jain.

A new front in the political battle

Deepfakes have opened a new front in the election battle between rival parties. The BJP has filed complaints with the Election Commission and the police regarding two deepfake videos related to Madhya Pradesh.

BJP’s IT Cell chief Amit Malviya has alleged that social media handles spreading the deepfake videos of KBC and CM Chouhan in Madhya Pradesh were related to the Congress.

“Congress is experimenting with it in MP polls, and will rely on it a lot in the run up to the 2024 Lok Sabha elections. Congress is looking to damage institutions, undermine people's confidence in democracy, and subvert their free will,” Malviya told India Today's News Director Rahul Kanwal on his program Newstrack on Thursday night.

The charge was refuted by Congress leader Rohan Gupta, who admitted that the cost of deepfakes is high. “If misused, even a small clip can cause damage by the time it is proved fake and taken down,” he responded.

What Internet companies are doing to protect elections?

Microsoft has announced measures for assigning digital watermarks for media content, which let users know how, when and by whom the content was created or edited and if it was generated by AI. It will also assist campaigns to deal with cyber security threats and measures for election watchdogs.

Meta, the company that owns Facebook and Instagram, has said that it would require political ads to disclose whether they used AI. Other unpaid posts on the sites that use AI, even if related to politics, would not be required to carry any disclosures.

What are the laws globally to stop AI meddling in elections?

Not many countries have well-defined legislation to deal with the threat of deepfakes in elections. In the US, many states have enacted specific laws. On the federal level, the Protect Elections from Deceptive AI Act was passed by the US Congress in September this year.

The European Union, through the AI Act, is working on regulating AI use. Proposed in April 2021 and still in negotiation as of mid-2023, it aims to classify AI systems according to the risk they pose to users. In the UK, the recently passed Online Safely Act regulates AI.