Deepfake Elections: How Indian Politicians Are Using AI-Manipulated Media To Malign Opponents

In poll-bound Telangana, hundreds of thousands of voters have received a forwarded video on their smartphones, which depicts a sitting minister appealing them to vote against the present state government.  

In Madhya Pradesh, videos have emerged that made use of clippings from popular TV show Kaun Banega Crorepati. The clippings show the Amitabh Bachchan-starrer quiz show asking questions around Madhya Pradesh politics to whip up anti-incumbency sentiments among viewers.  

In both cases, the videos depict events that never took place. They are deepfake videos, generated with the help of artificial intelligence (AI). After the rise of post-truth politics across the globe, which resulted in widespread polarisation of voters based on misinformation, the world is now entering an even more dangerous phase of deepfake politics.  

Election propaganda in India has evolved beyond door-to-door campaigns and wall posters to AI-generated fake videos. This technology allows a bunch of people sitting in their Delhi NCR offices to deploy deepfake videos that can sway voter sentiments in any poll-bound constituency hundreds of miles away.  

Deepfakes have been making headlines in India after a fake video of Telugu actor Rashmika Mandanna went viral earlier this month. IT Minister Ashwini Vaishnaw called deepfakes a “threat to democracy” and even Prime Minister Narendra Modi expressed his concern over the dangerous capabilities of the technology.  

But then, why aren’t political parties coming out hard against fake videos targeting their candidates? It’s because almost every major party has got their hands dirty in the business of deepfake propaganda. 

Political parties across India are employing private-sector advocacy and campaign agencies whose job is to further their client’s electoral prospects. In many cases, as illustrated above, this ends up involving deepfake videos as well.  

Outlook Business spoke to almost a dozen people working for political consultancies across the country, serving all major parties, to understand how deepfake media is being used in election campaigns to spread misinformation and manipulate voters’ perceptions.    

Misinformation Madness 

Deepfakes are misleading, synthetic media generated by artificial intelligence (AI), that make use of deep learning technology to generate real-looking images and videos. This technology allows for public figures to be shown as giving controversial statements, or engaging in problematic behaviour, even without their knowledge. 

“There are two main categories of deepfake videos that election campaigners deploy. One is aimed at creating a positive sentiment around the candidate we are endorsing, and the other is aimed at spreading misinformation about the opposing candidates,” says a political consultant who’s presently working on the Telangana state elections, scheduled for end November.   

It is understood that these deepfake videos are then shared across the voter base of the concerned constituency through WhatsApp groups. To spread misinformation at a hyperlocal level, groups are either created for microtargeting audience according to their demographic details, or sometimes groups are bought off from local social media influencers. The admin control over a group with 250 members from a single constituency is usually bought for close to Rs 5,000 in rural areas, says a consultant who’s worked in four state elections so far.   

Groups that are specifically created for the purpose of disseminating political propaganda, including deepfake-enabled misinformation, are known as ‘scratch groups’ within electoral campaigning circles.    

“Scratch groups are segregated into four types based on the degree of gullibility of group members. The age group between 18 and 25 is the most preferred since they can be influenced easily. The worst kinds of deepfake videos are shared in these groups, involving morphed images of female political opponents drinking alcohol in a private setting or of male political opponents who are depicted as womanisers or as irreverent to their religion and caste,” says the second consultant mentioned above.  

All the consultants and social media campaign workers quoted here requested anonymity due to fear of workplace repercussions, or harassment from political parties. 

The misinformation videos are largely designed to be shared only over WhatsApp, people working with the social media teams of multiple political consultancies admitted to Outlook Business. This is due to the fact that platforms like Instagram, Facebook and X provide users with the option to report suspicious and spam-like posts that spread misinformation.  

But since platforms like WhatsApp and Telegram are closed messaging platforms, the guardrails around content moderation are not identical to the ones followed by X and Instagram. “Eventually, it [deepfake video] gets shared on public facing social media platforms but by the time it’s reported and removed, it’s already circulated in hundreds of WhatsApp groups. That’s our target in the first place, Instagram videos or posts on X are only an after-effect of our work,” points out another political consultant who’s worked with two leading companies in this space.

All Hands Are Dirty  

It's not just India’s electoral system that is suffering from the menace of deepfake-enabled propaganda. In the US, the Federal Election Commission (FEC) already initiated a process earlier this year to potentially restrict all AI-generated deepfakes in political advertisements ahead of the 2024 elections. In the UK, the National Cyber Security Centre has already warned about the threat posed by deepfakes and other AI tools towards the next general elections in the country. 

India too is scheduled for national elections next year, preceded by the assembly elections across five states in November-December this year.  “All the parties are aware of how common these deepfake videos are. I deployed one [deepfake video] last week against our opponent and I know what the opposing candidates are doing as well. But nobody will take it public because everyone's hands are dirty,” says the consultant quoted earlier, who is working on the Telangana elections.   

In a meeting with social media companies held this week, the IT Ministry announced that it will develop new regulations around the detection of deepfakes, and prevention of its spread via social media. The Election Commission of India (ECI), however, has not announced any measures so far to reduce the impact of deepfake technology on the upcoming elections.  

S.Y. Quraishi, former chief election commissioner of India, says that the ECI has a formidable challenge on its hands given how quickly deepfake-induced misinformation spreads in the run-up to elections. ECI representatives did not respond to requests for a comment on this matter.    

The former ECI chief also cautions the election watchdog to act on deepfakes independently from the efforts at the IT Ministry. “Some of the deepfakes can come from the ruling parties also. So, although an alliance between the ECI and the IT Ministry sounds good on paper, there’s always the possibility of a collusion, or people in power keeping their eyes closed. So, it’s the ECI’s credibility at stake,” he points out. 

The Social Media Dilemma  

For social media companies, detection of deepfakes will be an expensive process that requires a significant amount of computing power. But, in case the platforms fail at controlling deepfakes, they stand to lose the ‘safe harbour’ provision that they enjoy under Indian laws, as minister Vaishnaw has clarified.   

This provision protects social media companies from regulatory liability on third-party content shared by users on their platforms. Since the provision is hard coded in the laws, only the judiciary can determine whether the companies can be held responsible for abetting unlawful content on their platforms.  

To protect themselves from prosecution, companies may have to invest in detection tools which might not be very helpful in the first place. “The detection process has many limitations because it’s looking for visual inconsistencies that suggest forgery. But the detection tools become less effective as the image quality of videos and pictures degrades,” says Meghna Bal, head of research at Esya Centre, a Delhi-based tech policy institute.  

For messaging platforms like WhatsApp, which is the primary conduit for political deepfakes, there is an added challenge of encryption. “A common proposal to preserve end-to-end encryption and restrict prohibited content is a method known as client-side scanning. Broadly, this involves pre-screening images before they are sent. The app checks the hash of the image or video against a database of hashes of prohibited content. It is unclear how this approach could work for deepfakes. Moreover, it still counts as a breach of privacy and could have a chilling effect on free speech,” explains Meghna.  

Under Section 4 (2) of the IT Rules, 2021, commonly known as the ‘traceability provision’, the government can ask messaging platforms like WhatsApp to disclose the original sender of a message. WhatsApp has challenged this provision in the Delhi High Court arguing that tracing a single message would require monitoring of all messages, and this would break the end-to-end encryption service it provides.   

For WhatsApp, the dilemma is on how to go about abiding the laws around deepfakes in India without ending the very service that allowed them to become a popular secure messaging platform.  

To The Polling Booth  

The next general elections to the parliament are expected to be held in April-May 2024. With less than six months to go, the election watchdog in the world’s largest democracy is yet to take any steps at curbing deepfake political propaganda.

In the run up to the previous general elections in 2019, the EC and social media companies had agreed upon a voluntary code of ethics to protect the free and fair nature of elections. The companies were represented by industry body Internet and Mobile Association of India (IAMAI). 

Notably, deepfakes were not mentioned in the voluntary code back then as the technology was yet to make its presence felt in the Indian electoral process. The number of smartphone users in India increased from a little over 500 million in 2019, to more than 650 million in 2023, according to Techarc data. This rise, along with the prevalence of low-cost, high-speed internet and cutting-edge analytics employed by political consultants, makes the spread of deepfake-enabled misinformation a very serious threat to India’s young electoral base.  

Now, as the country inches closer to what is expected to be its biggest election yet, deepfakes are spreading easily, one forward message at a time. With deepfakes eliciting strong reactions from the Prime Minister’s office and the IT Ministry, it remains to be seen whether the country’s election watchdog will also take up a strict position on the matter. If AI interference on the upcoming election is not curbed, the biggest election ever in the world’s largest democracy may well become the ground zero of large-scale deepfake elections.