Nivell Rayda, Jakarta – The frontrunner in the upcoming presidential election speaking fluent Arabic. A long-deceased president praising the incumbent. A presidential candidate being scolded by one of his political backers.

These audio and video clips have made the rounds on Indonesia's social media scene in the run-up to the country's Feb 14 elections, racking up countless shares through popular messaging platforms.

At first glance, they look convincingly real. But on closer inspection – with the help of a plethora of free websites and applications – they have been exposed as deepfakes, created or modified using artificial intelligence (AI).

Across the globe, the use of AI has been growing since the arrival of text-to-image AI models such as Dall E in 2021, and chatbots, including ChatGPT in 2022.

The global artificial intelligence market size was valued at US$196.6 billion in 2023 and is projected to expand at a compound annual growth rate of 37.3 per cent from 2023 to 2030, a study by Grand View Research suggests.

As its popularity grows, participants in the upcoming Indonesian presidential and parliamentary elections have dipped their toes into the world of AI.

They include the campaign teams of presidential candidates Mr Prabowo Subianto and Mr Ganjar Pranowo who told CNA that they are using AI to win the hearts and minds of Indonesia's 200 million eligible voters.

The other presidential candidate, former Jakarta governor Anies Baswedan, has not been using AI as part of his campaign, his team told CNA.

But analysts are criticising the use of AI for political purposes, warning that it can also lead to the proliferation of disinformation and black campaigns, political strategies which disseminate false accusations against opponents.

Even before the rising popularity of AI, disinformation has been widespread in Indonesia, most notably in the run-up to the 2019 presidential election which left the country of over 270 million people divided along political lines.

With AI, pundits fear, such disinformation could be more widespread and damaging.

Detecting and dispelling disinformation

The earliest example of politically-charged, AI-generated disinformation related to this election came in October last year when a video surfaced on YouTube showing the current president, Joko Widodo, popularly known as Jokowi.

The original video was taken from a 2015 speech the Indonesian president made at a gala dinner in the United States. Jokowi delivered the speech in English but the content was doctored using AI to make it appear that he was speaking Mandarin.

Although the one minute and 20 second video was immediately taken down, copies spread like wildfire on messaging services, particularly among people critical towards the president who are eager to portray Jokowi as pro-China and stir up the anti-China sentiment among conservatives and nationalists.

In late January, Mr Anies became the latest victim of disinformation when a video appeared on social media containing a voice recording of what appeared to be National Democratic Party (Nasdem) chairman Surya Paloh scolding Mr Anies over his performance at the polls and the recent presidential debates.

"Anies, your survey is always at the bottom. We party leaders are having a headache. The coalition is also having a headache. What do you say about this?" the purported Mr Surya said.

"I have tried my best. During the debate, I also worked hard to get people's attention," the supposed Mr Anies replied.

Nasdem is Indonesia's fifth biggest party in parliament and one of the early political backers of Mr Anies' bid for the presidency.

Experts pointed out that although the voices were similar to that of Mr Surya and Mr Anies, their choice of words, pronunciation and phrasing felt unnatural and out of character.

"We have to be critical, because now there is AI technology which can generate audio or visual (content) that can appear real," Mr Anies told a campaign rally in Yogyakarta province on Jan 23, shortly after the video surfaced.

Mdm Khoirunnisa Nur Agustyati, executive director of the non-profit Association for Elections and Democracy (Perludem) said that the government must act fast to dispel disinformation and stop it from spreading.

"Disinformation can be used to attack a particular candidate and this can lead to division in our society," she told CNA.

Mdm Khoirunnisa said even without the use of AI, disinformation was rampant during the 2019 election when Jokowi faced off with Mr Prabowo, leading to a deep political split which lasted even after Jokowi won the election by a margin of 11 per cent.

Now, with AI providing accessible tools for the spread of disinformation and political division, analysts are concerned it could be much worse.

The 2017 law on election bars a candidate from attacking an opponent. But unless officials can prove that these AI-generated black campaigns are created and spread by individuals with direct links to a campaign team the law cannot be applied.

"However, there are other laws which can penalise the perpetrators," Mdm Khoirunnisa said.

Among these regulations is Indonesia's information and electronic transactions law and the country's penal code which criminalises libel, slander and disinformation.

But not all disinformation is antagonistic in nature. Sometimes it can help boost a candidate's brand.

In November, two deepfake videos appeared around the same time on social media, one of Mr Prabowo and one of Mr Anies. Both were shown talking in fluent Arabic, a language neither of them speak.

The videos' appearance led many of their respective supporters to believe that their candidates of choice are intellectual multi-linguists with strong Islamic credentials.

Two months later, Golkar Party politician Erwin Aksa posted a video featuring the likeness of the late president Suharto, who died in 2008.

In the video, the AI-generated Suharto urges Indonesians to vote for legislative candidates from Golkar, which the late president established as a political party in 1964 and which served as his main political vehicle during his 32-year rule of Indonesia.

The AI-generated video then went on to praise Jokowi and his predecessor Mr Susilo Bambang Yudhoyono for carrying on Suharto's vision in developing the archipelago nation.

Both Golkar and Mr Yudhoyono's Democratic Party are supporting Mr Prabowo's bid for the presidency. Mr Prabowo's running mate, Mr Gibran is Jokowi's eldest son.

Mdm Titi Anggraeni, an election law expert at the University of Indonesia said the laws are vague when it comes to these types of disinformation.

"The laws only bar materials which defame another candidate, provoking civil unrest, (or which) contain threats of violence and so on.

These contents are indeed disinformation. But they are not defaming or attacking anyone. We must admit that the laws are not yet equipped to deal with them," she told CNA.

Mr Rahmat Bagja, chairman of the country's Elections Supervisory Committee (Bawaslu) said his office is working together with the Indonesian Ministry of Communication and Information Technology as well as various social media and messaging platforms to make sure that all forms of disinformation related to the upcoming elections are swiftly taken down.

"Bawaslu is hard at work monitoring election-related videos and deciding which are hoaxes and which are not," he told CNA.

On Dec 22, the communications ministry issued a circular to technology companies working on AI tools in the country to be mindful of the potential negative impacts of the products they are developing.

A regulation which governs how the average users should utilise this technology ethically and responsibly, is still being formulated.

How presidential candidates are using AI

The campaign teams for the three presidential candidates denied that they are using AI to spread disinformation which could either benefit themselves or smear their opponents.

Instead, those supporting Mr Prabowo and Mr Ganjar said they are using AI to provide new experiences for their supporters in a creative way.

Mr Prabowo's campaign team for example has been using AI to produce a childlike image of the presidential candidate, giving the 72-year-old retired general a softer and more approachable persona.

The chubby and cheerful cartoon character, nicknamed "gemoy", which means adorable, has been appearing in official banners, posters and t-shirts of Mr Prabowo's campaign.

"AI was used to create this image which does not reflect the real Prabowo. In the beginning people liked the new image, but this sentiment plummeted when people got to see Prabowo's true colours during the debate," social media analyst Ismail Fahmi told CNA.

Mr Ismail was referring to the three presidential debates staged during the campaign season which showed Mr Prabowo giving fiery remarks and appearing infuriated when he was criticised by his rivals.

The team has also created a generative AI tool which can make it appear that users are taking a selfie with Mr Prabowo and his running mate, Gibran Rakabuming Raka in a number of different backgrounds and poses.

"We need to keep up with technology," Mr Budiman Sujatmiko, an adviser at the Prabowo-Gibran campaign team told CNA. "We want to stay innovative in our campaign strategy and offer something new to the people of Indonesia."

Meanwhile, supporters of Mr Ganjar have been using AI to generate pictures of the 55-year-old former Central Java governor in trendy outfits, making the presidential candidate appear more youthful and hip. Mr Ganjar's team has used some of these AI-generated images as their campaign materials.

Both Mr Prabowo and Mr Ganjar have come up with their own chatbots to make it appear that users are chatting with their respective candidates.

"We use AI to learn how Ganjar might respond and users can have their questions answered the way Ganjar might answer them," Mr Pangeran Siahaan, a spokesperson for the Ganjar-Mahfud campaign team, told CNA.

Mr Anies on the other hand prefers not to use AI as part of his campaign strategy due to its artificial nature, said Mr Angga Putra Fidrian, the Anies-Muhaimin Iskandar campaign team spokesperson.

"(Mr Anies) is more in favour of real life dialogues with his constituents and letting the public see him for who he is," Mr Angga told CNA.

Mr Ismail questioned the effectiveness of these AI-based campaign strategies arguing that the artificial nature of AI is the technology's pitfall.

"At first, there was this hype over AI and everyone was racing to find ways to use it. But over time, people are getting tired of it because AI cannot represent the truth. It cannot represent a person, the person's idea and complex character. It doesn't substitute human engagement," he said.

Instead, the arrival of AI is keeping election officials, campaign teams and civil society on edge about its potential to mislead the public, Mr Ismail continued, whether it was intentional or not.

Chatbots developed by the Prabowo and Ganjar campaigns, for example, have the potential to stretch the truth.

"They were created so people can ask about (the candidates') programmes. But you can't tell what people will actually ask," he said.

Mr Ismail said that he once tried to ask one of these chatbots about the seventh principle of Pancasila, even though the Indonesian state philosophy only has five principles.

"The chatbot erroneously came up with the sixth and seventh principles even though they don't exist. Things like this can ruin a candidate's credibility," he said.

Mr Ismail said that AI is also empowering the so-called "buzzers", internet trolls who get paid to generate content for political influence on social media and news websites.

"Before (AI), buzzers needed to manually formulate their own posts or comments. With AI, all you have to do is input a prompt and it can generate hundreds of responses in a matter of seconds," he said.

Ammunition in the battle of ideas

But AI can also help streamline the election process.

The General Elections Commission (KPU) office in Sukoharjo Regency, Central Java, told local media in December that they are planning to use AI to expedite its vote tallying process.

Although the official result of the presidential and parliamentary elections will still rely on manually counting the vote on each ballot paper, data processed by the technology, the office said, will help identify if there are errors and anomalies.

Other KPU offices, including its headquarters in Jakarta, have not indicated that they will follow the Sukoharjo office's plan.

Aside from choosing the country's next president, more than 204 million Indonesians will hit the ballot booths on Feb 14 to also decide on members of its national, regional and city-level parliaments.

In total, there are more than 20,000 positions up for grabs, highlighting the complexity of what election officials will face on the ground once the votes are in.

"There are more than 300,000 people aspiring to be members of parliament. Not everyone has access to data and information. Even if they do, not everyone has the ability to analyse this data and information and translate them into a campaign strategy," entrepreneur Jose Rizal told CNA.

This is one of the reasons why Mr Jose started Pemilu.AI in April last year.

The firm is using AI to analyse publicly available information such as those from the Central Bureau of Statistics as well as qualitative data from face-to-face interviews so that legislative candidates can identify issues relevant to their constituents, formulate a campaign strategy and come up with a detailed execution plan.

"With the help of AI, we can also identify what demographics the candidate should be targeting. AI can also formulate ways to attract these target groups from activities they might find interesting all the way to campaign slogans and social media posts," Mr Jose said.

So far around 700 parliamentary candidates are using Pemilu.AI's services, the firm's founder said, spending IDR 29 million (US$1,838) for every city and regency level candidates, IDR 99 million for provincial level MP hopefuls and IDR 149 million for national level parliamentary candidates.

Although the cost is a small fraction of what candidates might spend if they were to go to a traditional political consultant, AI technology still has its limitations.

"Can AI really substitute the role of consultant? Of course not. Humans have intuition and experience. AI can conceptualise a strategy but its execution still depends on the candidates, how they can get their message across and how they interact with their constituents," Mr Jose said.

Pemilu.AI is not helping presidential candidates because it would mean collecting vast amounts of data from every corner of the world's fourth biggest nation in population.

However, the campaign teams of all three presidential candidates admitted to CNA that they too are using AI to map out their target voters, formulate their respective campaign strategies and monitor their effectiveness.

The three teams refused to divulge further about the AI-based tools they are using and what technology consultants they are working with.

Mdm Titi said it is more fitting for AI to be utilised as means to help election officials and campaign teams perform their tasks better as opposed to being used for gimmicky and superficial things like campaign posters and chatbots.

"An election is supposed to be a battle of ideas," she said. "AI or any other technology shouldn't be used as gimmicks, for attacking others or creating disinformation.

We need this election to have class because an election is meant to pave the way for the creation of a civilised society." – CNA/ni(ao)

Source: https://www.channelnewsasia.com/asia/ai-disinformation-deepfakes-indonesia-elections-409129