Indonesia Elections: A GenAI Race To Win GenZ's Hearts
Debriefing Indonesia Elections - More about AI's attack patterns - Talking about the Blue Framework
Hey there.
This week has been filled with two important elections, Pakistan and Indonesia. Two elections that are particularly revealing in terms of information manipulation activities, including various uses of generative AI (GenAI).
We discover both cute ways and more gloomy ways to use AIs to shape voters’ perceptions. We also discuss whether AI is an expansion of mass media or a new kind of media that builds bridges between close and open spaces, as well as between the past and the present.
As the election cycle continues, many also discuss solutions and countermeasures, also known as the Blue Framework. These are listed for your own reflection on their value and downsides.
Let’s start!
Indonesia Elections: Setting The Scene
Yesterday, Indonesia held its presidential and legislative elections, marking a crucial moment for geopolitical players for several reasons.
First, in terms of demography, Indonesia is the world’s third-largest democracy with nearly 205 million eligible voters. It is only the 5th time Indonesian are casting their votes in a presidential election since Indonesia started its democratic reforms in the late 1990s and its first direct presidential election in 2004. This election is therefore critical to consolidate the democratic Indonesian institutions, especially as former President Joko Widodo is stepping down after two five-years term, and criticism that democratic norms have eroded during his leadership. There are also fears that President Widodo might seek to retain its influence, even after the end of its mandate. His son, Gibran Rakabuming Raka, has emerged as the leading candidate for vice president. Pretty good for a former leader of a catering business and a chain of dessert shop, right?
Secondly, in terms of geopolitics, Indonesia holds high stakes for both the US and China due to its regional power. The political stability of Indonesia directly influenc the stability of the entire region. Throughout President Joko Widodo’s terms, Indonesia maintained a neutral relationship between the two superpowers, but it might change in the future. The three candidates - Anies Baswedan, former governor of Jakarta, Prabowo Subianto, current defense minister and former general in the Indonesian military during Suharto’s regime; and Ganjar Pranowo, former governor of Central Java - have all expressed their will to maintain Indonesia’s “free and active” foreign policy. However, there are slight variations among the three candidates.
Baswedan embraces a multilateralism approach, aiming to position ASEAN as a central forum for ensuring regional stability.
Subianto supports a “good-neighbour” policy, emphasizing the cultivation of good relationships with all significant players.
Pranowo envisions Indonesia as a more proactive player in international affairs, seeking to modernize the “free and active” foreign policy of the country.
Thirdly, Indonesia is a central economic power with the potential to influence the future of environmental policies. It is one of the world’s biggest exporters of coal, nickel and palm oil. This economic strength has come at the expense of Indonesia’s environment. Forests have been razed, coastal waters and waterways polluted, and cities affected by smog. For both external and internal actors, the choice of a presidential candidate will reflect Indonesia’s future, currently torn between fast growth and a healthy environment.
Who Can Influence the Outcome of Indonesia Election?
The choice of a new president can be influenced by two types of actors: foreign actors, notably China and Russia, and domestic actors, such as extremist and terrorist groups.
Foreign actors, such as China and Russia, pose credible threats that could influence the election’s outcome.
Regarding Russia, I was initially skeptical that it might be a threat actor targeting Indonesia. However, there has been reports, that in the previous 2019 presidential and legislative elections, Russia sought to undermine the elections through cyberattacks targeting the voters’ list. According to the Center for Information Resilience, a popular Indonesian YouTube channel ‘Tribun Timur’ has been spreading pro-Russian narratives and official Russian government content to glorify the Russian army while undermining the Ukrainian army. According to the ASPI, in the context of the Ukrainian war, Russian propaganda, leveraging anti-Western sentiment, has shaped public opinion in Indonesia.
When it comes to China, the balance of power within the region, the economic competition between the US and China, and the neutral policy of Indonesia create a plausible scenario for potential Chinese interference in the elections. The influence of the People’s Republic of China (PRC) is already expanding through media.
The PRC employs recurrent tactics, developing content-sharing or content-generating agreements with local media outlets or sponsoring local journalists’ trip to cover PRC events. These tactics aim to extend the reach of PRC’s narratives within the country. These agreements and trips often result in the absence of criticism of the PRC and a blurred line between the media’s editorial stance and the PRC’s own content.
These two foreign actors may have taken steps to increase their information manipulation campaigns to influence the turn-out of the election. They may also be responsible for this online hate campaign turning Indonesians against Rohingya refugees. According to the UNHCR, there is a “coordinated online campaign of misinformation, disinformation and hate speech for refugees”, which started just months before the election and looks professionally made on Instagram, X and TikTok.
This type of campaign might not have been planned by a foreign actor but could become useful to amplify local fears and anxieties which could turn into voting for a more authoritarian and populist figure.
It could have been spread by extremist groups, who have been active in the country for some years. In 2017 and 2019, according to the Global Network on Extremism and Technology (GNET), two incidents disclosed an increasing polarization within society, between Muslim Indonesians and Chinese Indonesians.
Terrorist groups such as the Indonesian branches of the Islamic State (JAD) and Al-Qaeda (JI) have also played a role in influencing elections. The GNET underscores how a past violent history, constant hatred between groups and modern conflicts are being used to shape an imaginary full of fears and distrust within the Indonesian population.
Targeting The Youth: The Rise of TikTok & AI
These vulnerabilities are being exploited by foreign or domestic actors to influence Indonesian voters, with a specific focus on one particular audience: the youth. Millennials and Gen Z make up 56,5% of the electorate, holding particular concerns about the environment and the economy. This audience is predominantly present on one platform: TikTok. Did you know that Indonesia is TikTok’s second biggest market?
On this platform, candidates have employed all kinds of strategies to connect with young audiences through youth-oriented content.
Bye bye long boring national speeches.
Hello tears, penguin symbols and K-pop videos.
The youth audiences appears particularly drawn to emotional content and authentic personal views, a type of content amplified by TikTok’s algorithm.
In this context, one may wonder whether political candidates are truly the ones in control of their campaign or if voters are influencing them through their local digital practices?
Generative AI is also not at rest here and is actively employed by political candidates to engage with young voters. For instance, Prabowo Subianto has rebranded his campaign with the assistance of the U.S. firm Midjourney Inc. GenAI was used to produce a doe-eyed cartoon version of Subianto, triggering 19 billion views on TikTok and appearing on sweatshirts and stickers.
Gen AI Cartoons of Prabowo (L) and Gibran (R), the son of current President Widodo.
One can observe that Prabowo attempts to present himself as a “cuddly grandpa” to appeal to his youngest audiences. Cute, isn’t it?
Well, less cute when one discovers that he was responsible for the abduction and killing of pro-democracy student activists during Suharto’s regime. A part of history that young Indonesian audiences do not learn at school, as it is omitted from textbooks to conform with the current politics of memory in Indonesia.
Gen AI is not confined to Prabowo and the generation of cartoons. According to Katie Harbath, who has a long history of working at the intersection of elections and tech, it is the first election where these tools are being used extensively. All candidates are leveraging them to track social media sentiment, develop chatbots, and target voters. Among these tools, one app seems particularly successful :“Pemilu.Ai”. Allegedly, it has provided services to 700 legislative candidates, enabling them to craft hyper-local campaign strategies and speeches. This showcases the economic opportunities for the private sector to sell GenAI’s services during elections.
However, these digital aspects should not let us forget about the other part of the Indonesian population who is unfamiliar with writing instruments. 34% of Indonesian are still not online.
As the New York Times reports it, voters will “mark their ballots by hammering nails into them, which election officials say is a fairer method than using a pen”.
To maintain the fairness of the vote, at least online, several actors have contributed to countering information manipulation activities. Local factcheckers, mostly volunteers, actively debunk misleading information related to elections. Companies like TikTok and Meta are paying fact-checking organisations such as AFP to verify videos.
Following the Path of AI
No, AI has not become my favorite topic, but before we move on to another election, it might be worth noting two persistent attack patterns involving GenAI that have serious consequences for elections.
Last week, reports emerged of Iran’s latest cyber-enabled information operation targeting the United Arab Emirates. Analysts at Microsoft have attributed the campaign to Iranian actors. It involved interrupting TV streaming services in the UAE to broadcast an AI-generated anchor delivering a report on the war in Gaza.
Using AI to impersonate news anchor is becoming a recurrent attack pattern, employed primarily by state actors,= such as Russia, China and now Iran.
In 2019, Russia, with the support of China, created the first Russian-speaking GenAI news anchor. Last year, Chinese actors used GenAI to create fake English-speaking news anchor targeting U.S. audiences.
This is a reminder that news anchor are significant targets for threat actors, not only to create new ones but also to leverage existing ones, as it was the case last October when a deepfake video impersonated a VOA Russian-speaking news anchor to promote a trading product.
Another recurring attack pattern is much gloomier.
It involves the resurrection of dead politicians to exploit their past authority and prestige to spread political messages to voters. This has already been the case this year in Indonesia and India. This raises questions about the ethical and legal framework surrounding the use of AI to counter its malign uses.
These two attack patterns raise questions about the impact of GenAI on influencing voters compared to other media. Is GenAI a continuity and amplifier of mass media or is it a new kind of media breaking down previous barriers?
The election last week in Pakistan may give us a hint.
The election ended up without a clear winner, amid a chaotic situation marked by internet shutdowns, platforms outages, and one political candidate campaigning from his jailhouse. No party managed to secure enough seats, but imprisoned ex-Prime Minister Imran Khan’s party won the most seats in the lower house of parliament. As you may recall, the party had used GenAI to convey Imran Khan’s speeches from the jailhouse to his online supporters.
This outcome demonstrates how GenAI can assist candidates and parties in circumventing restrictions in a particular electoral context. It also highlights the limits of GenAI.
In the case of Pakistan, an authoritarian regime, having no clear winner at the end of the election means that only a battle for narratives can resolve this issue, imposinge an authoritative win without democratic support.
Your press corner
Here’s the weekly readings to keep you connected to all the conversation on global elections and information operations:
Chinese firm behind ‘news’ websites pushes pro-Beijing content globally, researchers find | Reuters - More than 100 websites disguised as local news outlets in Europe, Asia and Latin America are pushing pro-China content in a widespread influence campaign linked to a Beijing public relations firm, digital watchdog Citizen Lab has found.
China’s YouTube Propaganda in Latin America – The Diplomat - China also seeks to influence audiences in Latin America, in an influence battle with Russia and Iran.
France accuses Russia of a disinformation campaign in a key election year | AP News - about the latest Russian operation discovered by VIGINUM.
Steve Bannon hawks disinformation to support Trump as legal troubles mount | Steve Bannon | The Guardian - Here we go again… Steve Bannon is pushing a “tidal wave of election disinformation” on his War Room podcast in the run-up to the U.S. election.
AI and misinformation: what’s ahead for social media as the US election looms? | US elections 2024 | The Guardian - a good summary of where we AI stands.
The ASD AI Election Security Handbook – Alliance For Securing Democracy (gmfus.org) - ASD released its AI Election Security Handbook last week.
The election in India is playing out on YouTube - Rest of World - When politicians turn themselves into YouTube influencers.
How X Is Trying to Win Over Influencers - The New York Times (nytimes.com) - Linda Yaccarino, X’s chief executive, is now relying on her television industry ties to make the site a destination for video creators. This could completely change the nature of the former Twitter platform.
The rise of the TikTok news anchor (economist.com) - our new generation of online influencers.
GIJN’s Elections Guide for Investigative Reporters — Revised for 2024 – Global Investigative Journalism Network - The Global Investigative Journalism Network reporting guide, updated and revised for 2024, is “designed to offer a broad array of tools, techniques, and resources — beyond the primary local sources you find — to help watchdog reporters dig into almost any campaign or election.”
Keeping it authentic: the social footprint of the trolls’ network | Social Network Analysis and Mining (springer.com) - based on the argument that - because Russian trolls need adhere to social expectations to pretend to be legitimate social actors, they leave a social footprint - this research suggests that building a machine learning model around social functions can be used to map and identify Russian influence networks.
How Much Do Re-shares on Social Media Change Beliefs or Opinions? - R Street Institute - apparently not so much.
European Commission Seeks Public Input on Proposed Digital Services Act Guidelines to Enhance Election Integrity | TechPolicy.Press - in case not seen, for those who want to influence the DSA guidelines.
Disinformation Isn’t Just a Problem for Elections – Foreign Policy - a good reminder that threats posed by social media, equally pressing, are not limited to electoral context.
Avoiding the Disinformation Trap | The New Yorker - a very interesting article about the backlashes of using terms such as “disinformation” and “fake news” to counter this threat.
And just for fun : West to use online platforms to influence Russian presidential elections — diplomat - Russian Politics & Diplomacy - TASS
Talking About the Blue Framework
I wanted to finish this newsletter by taking the time to draw a picture of recent countermeasures that have been taken to address information manipulation activities:
France24 has reported on the admission by US right-wing group Project of its amplification of inauthentic electoral fraud claims in Pennsylvania during the 2020 U.S. election. France24 highlights that defamation suits may have an impact on deterring the spreading of electoral fraud narratives, but it may not be enough to change the voters’ beliefs.
The New York Times wrote about the Brazilian federal police operation involving search and arrest warrants to investigate former President Bolsonaro and his aides’ activities to hold power after the Brazilian elections in 2022. While these measures can support accountability, there can fuel another narrative such as the one spread by former President Bolsonaro that “he was the innocent victim of a politically motivated operation”.
WGAL, Pennsylvania, has reported about the U.S. Federal Communications Commission (FCC)’s new rule to ban GenAI robocalls. According to interviewed experts, the measure could help to discourage legitimate entities from using this deceptive tactic. But others may not be deterred at all.
Radio Taiwan International wrote last week about the need for fact-checkers to rely on assistance from AI tools to keep pace with the spread of online information manipulation. I personally attended a course by the Knight Foundation on AI in the newsroom which I found highly interesting to understand how AI can support journalist and fact-checkers work.
CSO reported about the US Cybersecurity and Infrastructure Security Agency (CISA) new measure to deploy additional regional inspectors to secure this year’s election. Is it a question of resources or a stratcom measure to deter malign cyber actors from interfering in U.S. elections?
Thank you for taking the time to dive into this newsletter and let me know what you thought about it and tips to improve it!
Ps: I wanted to talk about VIGINUM’s latest report on Portal Kombat, but the newsletter was already too long. It will be featured next week to discover all the vertiginous clues that the Internet infrastructure can reveal when one knows where to look at.