AI generated images were widely shared on social media platforms in the run up to the recent legislative assembly elections in Telangana. While this may be a prelude of what to expect in the upcoming Legislative Assembly and General Elections in 2024, there’s very little being done by tech companies to safeguard their system. The safeguards of image generation tools are easy to bypass and the guidelines don’t seem to have any impact. Meanwhile, even as these AI generated images went viral on social media, the Election Commission has stayed mum throughout.
AI generated media on social media
AI generated images of K. Chandrashekhar Rao of the Bharat Rashtra Samithi were widely shared on social media platforms including X (formerly Twitter) and Instagram in the run-up to polling in Telangana state. While a majority of the images shared can be characterised as simply glorifying the subject candidate, some of the images that were identified can be characterised as being misleading.
AI generated Images of K. Chandrashekhar Rao having breakfast with school children
For instance, the above set of AI generated images were shared by a user on X with the misleading caption "for those who missed KCR sir at the launch of [the] CM breakfast scheme." The post itself did not provide any indication that the images were AI generated. The scheme referred to in the post was launched by the Telangana government on 6 October, 2023 with the aim of providing breakfast to students of certain government-run schools in the state, however, K. Chandrashekhar Rao was not present at the launch for the scheme, and no photographs exist of K. Chandrashekhar Rao sharing a meal with school children in relation to the scheme.
AI generated Images of K. Chandrashekhar Rao meeting with dairy farmers
In another case, the above set of AI generated images were shared by an Instagram account with more than 50,000 followers with the hashtag "#KCROnceAgain”. AI generated images of Revanth Reddy of the Telangana Congress as well as Rahul Gandhi of the Indian National Congress were also shared on Facebook and Instagram as recently as a day before polling in Telangana.
AI generated Images of Rahul Gandhi and Revanth Reddy
These images, similar to the others highlighted above, illustrate a trend of AI generated images being used to present a particular candidate in a positive light – albeit by depicting fictitious situations and occurrences – a far more concerning trend was observed with other images that were shared on social media platforms in the run up to the Telangana elections.
The Telugu Desam Party did not contest in the legislative elections this year -- notably, for the first time since the party was formed in 1982 -- due to party frontrunner and former Andhra chief minister N. Chandrababu Naidu having been arrested and jailed in connection with a corruption case. And while a real photograph of N. Chandrababu Naidu in Rajahmundry Central Prison does exist, several AI generated images showing the politician behind bars were found to have been shared on social media.
AI generated Images of N. Chandrababu Naidu in prison
How were these images generated?
One of the accounts that originally shared the AI generated images of K. Chandrashekhar Rao and N. Chandrababu Naidu stated on X that the images were created using Microsoft’s Image Creator platform. Image Creator is an AI powered image generation platform that uses OpenAI's DALL-E 3 text-to-image model for generating images and is available to the public for free.
The terms of service for both OpenAI and Microsoft prohibit the use of their services to generate images resembling public figures, celebrities, and politicians, among other restrictions. The content policy for Image Creator also restricts “content that is intentionally deceptive and likely to adversely affect the public interest (including deceptive or untrue content relating to health, safety, election integrity, or civic participation) and inauthentic interactions.” These policies are enforced by technological measures meant to limit the creation of content that may violate terms of service.
One measure is a filter to check the textual prompt used to guide the generation of images. If a request to generate an image contains words or references that might violate the platform's terms of service, the request is stopped immediately and no images are generated. The other measure is meant to act as a fail-safe mechanism in case the first one fails, where generated images are visually checked to detect whether or not they contain prohibited material before they are shown to the user.
While these restrictions are effective at times, users have found ways to trick platforms like Image Creator into generating images that violate their terms of service. In September 2023, a Reddit user shared a post on the DALL-E 2 subreddit describing a technique to bypass Image Creator’s filters and generate images that would otherwise be blocked. Users swiftly responded with images they generated of American public figures and politicians using the technique mentioned in the post.
A Microsoft spokesperson told Decode that “Image Creator is a tool designed to help inspire people’s creativity. As with any new technology, some are trying to use it in unintended ways. When these reports surface, we take immediate action to address them in accordance with our content policy, and any repeated attempts to produce content that goes against our policy guidelines may result in loss of access to the service.” The spokesperson further stated “We have large teams working on the development of additional techniques and safety systems in line with our responsible AI principles and remain committed to continuing to refine our tools for the year ahead.”
While trying to verify the effectiveness of Image Creator’s safeguards, Decode discovered a prompt bypass that successfully generated images of Indian Prime Minister Narendra Modi.
AI Generated Images of Prime Minister Narendra Modi
Though Image Creator refused to generate images with the prompt “PMO India” or “Narendra Modi,” submitting the prompt “a respectful meeting between two influential leaders of India” consistently resulted in images being generated of Prime Minister Narendra Modi meeting with seemingly fictitious individuals. We reported this prompt to Microsoft using the company’s “AI-Powered Feature Reporting Form.”
It is important to highlight that not all AI image generation platforms even have guidelines prohibiting the creation of sensitive or deceptive images. A simple Google search led us to several platforms without a published content policy or terms of service in place, as well as several other platforms which blatantly advertised themselves as being “uncensored” or “unfiltered” alternatives to popular tools.
Besides, considering that Stable Diffusion – the technology powering most popular AI image generation platforms – is freely available, anyone with technical knowledge and computing resources required to set up the tool would be in a position to create AI generated images without being bound to any of the restrictions of popular platforms.
The response of big tech
The Economist has called 2024 "the biggest election year in history." With elections poised to be held in several major countries including the United States, United Kingdom, India, Mexico, among others – and with the misuse of AI generated images and deepfakes not being a problem that is specific only to India – several companies have identified the threat that synthetic AI generated media poses to the global electoral landscape and announced plans to address the issue.
Meta announced in November 2023 the introduction of a new policy for Facebook and Instagram where electoral or political advertisements as well as advertisements related to social issues would be required to disclose whether or not they make use of imagery that has been altered or wholly created by AI. Meta states the policy will be enforced globally from 11 January 2024 onward, and that they will "add information on the ad when an advertiser discloses in the advertising flow that the content is digitally created or altered."
Decode reached out to Meta for a comment on AI generated images of politicians that are widely shared on their platform. Meta did not respond to the query at the time of publishing this story.
Microsoft announced in November 2023 that it will be taking a number of steps to protect the integrity of elections worldwide. In a blog post for the announcement, the company stated that in 2024 the world "may see multiple authoritarian nation states seek to interfere in electoral processes" and that such interference "may combine traditional techniques with AI and other new technologies." In response to this threat, Microsoft says that it will provide services to political campaigns around the world, including making available “Content Credentials as a Service,” which combines metadata, digital watermarking and cryptography to allow users to provide details about the provenance of the images or videos they share. Microsoft will also deploy a Campaign Success Team to help "political campaigns navigate cybersecurity challenges and the new world of AI."
Ultimately, it is unclear whether these measures adopted by Internet companies will be helpful in preventing the political weaponization of generative AI technologies. India’s Ministry of Electronics and Information Technology issued an advisory on 26 December 2023, reminding intermediaries to ensure compliance with the country’s IT Rules, specifically highlighting the threat posed by deepfakes and AI generated misinformation. While this is a step in the right direction, India’s electoral landscape may certainly benefit from the Election Commission of India framing guidelines regulating the use of generative AI technologies in political campaigning.
Further, adoption of legislation similar to "Protect Elections from Deceptive AI Act," which is a bipartisan Bill that was tabled in the United States Senate in September 2023 that is meant to restrict the distribution of “materially deceptive AI-generated audio or visual media prior to election” would also stand to benefit the electoral landscape in India.