The meteoric growth of artificial intelligence (commonly referred to as AI) has created many opportunities globally, it also raise profound ethical concerns. The United Nations in its Interim Report: Governing AI for Humanity (the Report) has called for a closer alignment between international norms and how AI is developed through recommendations to enhance accountability. The Report highlights the importance of identifying, classifying and addressing the risks emanating from AI based on the principles of inclusivity and public interest. AI technologies have capabilities to generate and disseminate false information at an alarming scale and speed. With several economies having elections in 2024 including Bangladesh, India, UK, European Union, US, Pakistan, Mexico, Indonesia, Russian Federation among others, the concerns of AI influencing elections has triggered the nations to take preventive steps.
AI creates a complex web through which it can undermine the sanctity of the ballot box, manipulate voter perceptions, carry out micro-targeting propaganda, as opposed to the overt threats of cyber-attacks, on election infrastructure.
During the US presidential elections, the data science firm Cambridge Analytica rolled out an extensive advertising campaign to target voters based on their individual psychology. As a result different voters began to receive different ads based on their susceptibility to perceptions and arguments. This was possible because of the availability of real-time data on voters based on their behavior on social media, consumption patterns and relationships. These digital footprints were then used to build unique behavioral and psychographic profiles of cyber-citizens that could directly target them.
In 2024, traditional ways of conducting elections have completely transformed. Earlier elections meant campaigns involving physical rallies, slogans, and ballot papers. With time, ballot papers transformed to Electronic Voting Machines, and physical campaigning into social media posts, videos, and extensive media coverage among other things.
AI redefining the elections systems:
AI powered bots and tools are used by political parties to monitor engagement of social media content and posts in real-time.
AI technologies, such as voice recognition and natural language processing, are used to make voting more accessible to people with disabilities by enabling them to vote independently using assistive devices.
Predictive analysis through AI can forecast the potential impact of various campaign actions from advertising placements to public appearances and policy announcements.
Election authorities can use AI to monitor and identify violations of campaign laws, finance laws and ensure compliance with electoral regulations. In 2021, the Bihar Election Commission tied up with an AI Firm Staqu to use video analytics with the optical character recognition (OCR) to analyse CCTV footage from counting booths during panchayat elections. Drones are being used to monitor election events and rallies.
The misuse of AI impacting various stakeholders:
The integration of AI into election processes brings forth several critical issues that need careful consideration:
Data analysis and voter targeting – AI analyses vast amounts of data to identify voter preferences, behaviors, and demographics to target specific voter segments more effectively. A popular search engine, in 2005 added a program for targeted advertisements which personalizes content depending on users’ assumed interest. Resultantly, voters were forced to absorb content that consciously created a sub-conscious impact!
Spreading misinformation and deepfakes – The World Economic Forum’s Global Risks Report 2024 ranks misinformation and disinformation as the number one (1) risk for next two years. AI based dissemination of misinformation is like a tsunami - strong, wide and literally unstoppable. The recent deepfakes videos of Bollywood actors Amir Khan and Ranveer Singh, the video of Moldova’s president, audio clone of the US president of the US are examples of deepfakes interfering with elections.
Media’s role in electoral systems today cannot be undermined due to its expansive outreach and accessibility amongst the masses.
Bias and fairness - AI systems can perpetuate and exemplify biases present in the data they are trained on, disadvantaging certain demographic groups, thereby integrating bias in the electoral process, such as voter registration, candidate selection, or even result prediction. A classic example of AI perpetuating bias is where Google’s AI platform suggested that the policies of the Indian Prime Minister are “fascist” which led to social media debates on the political propaganda to create influenced narratives.
Weapon against women – In 2023, a manipulated image of women wrestlers was spread showing them with wide smiles while protesting against the President of the Wrestling Federation (while in the original photo they looked serious) thereby discrediting the wrestlers. Deepfake videos of female opposition politicians wearing objectionable clothes emerged in Bangladesh ahead of elections, is another example of targeting the integrity of a certain vulnerable social groups to the advantage of others.
Geo-fencing – Geo-targeting industry obtains location data from online ad placement system known as “real-time bidding” in which publishers or app owners share device, ad unit and geo-location information to a split second auction. In 2018, a conservative political party reportedly used location data to target people who frequently attended Catholic churches in Missouri over a period of 60 days. This gives political parties and miscreants an added advantage and an upper hand over voter preferences.
Non-consensual use of data – Utilization of AI in election campaign raises ethical concerns revolving around manipulating voters’ perceptions using personal data and behavioral patterns through data that is already in public domain where the legal liability would be difficult to assign.
Foreign interference – Russia was accused of interfering in the 2016 US elections, which was believed to have impacted electoral outcomes. The rise and availability of advanced AI-based sorting tools have enabled party arms and State security apparatuses to sort colossal amounts of data into precise and actionable information down to the individual. This can lead the political parties a fatal win or defeat!
Integrity - Days before the elections in Slovakia, an audio recording surfaced on Facebook, allegedly capturing a conversation between a candidate and a media representative discussing plans to manipulate election, including buying votes.
In a deepfake video, one member of a major political party was seen encouraging people to vote for his son who is the candidate for that political party in the elections. Even though he had passed away four years ago, his likeliness was reproduced using AI tools.
A very popular Common Man’s Party in Delhi used ‘blasters’ in which AI generated voice of the candidate addresses and greets the recipient of the call by their name.
How can deepfakes impact election results?
False narratives with widespread outreach can completely sway public opinion in the run-up to the elections which can cause questionable dents to the democratic landscape of any country.
Responsive Actions
The social media intermediaries have taken the following mitigating steps:
Meta had announced that it would be launching a new helpline in collaboration with the Misinformation Combat Alliance (MCA) which aims to “combat media generated using artificial intelligence."
WhatsApp has rolled out a new feature for users in India to report suspicious content through multilingual helpline chatbot in the app, supporting English, Hindi, Tamil and Telugu Languages.
Google had also announced collaboration with various major technology companies to help prevent deceptive AI-generated imagery, audio or video content from interfering with the global elections.
Logically.ai is one of the world’s largest, commercial fact-checking organisations which facilitates the identification of inauthentic information/ content prevailing in the public space.
In 2021, Adobe, Microsoft, Intel, Arm, BBC and Truepic had launched a coalition for standards development. The Coalition for Content Provenance and Authenticity (C2PA) is a consortium to accelerate the pursuit of pragmatic, adoptable standards for digital provenance, publishers, digital creators, media platforms and consumers.
The European Union’s Artificial Intelligence Act, 2024 (The EU Act Act)
The EU AI Act classifies AI systems as ‘High risk’ category based on its intensity of impact and prescribes transparency requirements to be complied with and imposes the mandate of labeling on AI generated content.
The United States
Michigan, in November 2023, has become the fifth state after California to regulate the AI in election communications punishing anyone who knowingly circulates an AI-generated deepfake within 90 days of an election.
The Guidelines Issued by the Federal Election Commission of America
The Federal Election Commission of America has issued guidelines, for political parties for making any public communication, that mandates displaying a “disclaimer” which should be clear and conspicuous.
India
The MEITY is likely to amend the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021(as amended on April 06, 2023) to include rules eliminating AI bias, to check and prevent deepfakes and to regulate fraud loan apps on social media.
The government has also notified the Digital Data Protection Act, 2024.
However, the proposed amendments and the Digital Personal Data Act, 2023 are only to be notified after the Lok Sabha elections. To fill the vacuum, MEITY has issued advisories stating that the existing law is well-equipped to deal with deepfakes which are made actionable under the Indian Penal Code 1860 along with the IT Act, 2000 and the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021;
The Election Commission of India (ECI) has issued the Guidelines titled ‘Responsible and ethical use of social media platforms and strict avoidance of any wrongful use by political parties and their representatives during MCC period in General Elections and byelections-regd’ dated May 6, 2024 directing political parties to take down deepfakes within three hours.
The ECI on April 7, 2024 issued another Press Note wherein, while urging the young and first time voters to participate in the elections through its ‘Turning 18 campaign’, it also highlighted the proliferation of fake news and misinformation online.
The National Cyber Crime Reporting Portal further allows cyber-citizens (netizens) to report the incidents of cybercrime including deepfakes.
As the world grapples together with the challenges posed by rapidly developing technology, regulating AI has become imperative to ensure the truthfulness and credibility of information that is made available to the public. With EU putting forth the Artificial Intelligence Act, the path is set for other countries to follow.
About the authors: Vikrant Rana is the Managing Partner of S.S. Rana & Co. Anuradha Gandhi is a Managing Associate and Rachita Thakur is an Associate at the Firm.
If you would like your Deals, Columns, Press Releases to be published on Bar & Bench, please fill in the form available here.