Perspective: Combating Deepfakes | 04 Jan 2024

What is the Context?

Deepfake has emerged as a serious threat to democracy and social institutions across the world. Propagation of deepfake content via social media platforms has aggravated this challenge.

• The Ministry of Electronics and Information Technology has, from time to time, advised social media platforms to exercise due diligence and take expeditious action against deepfakes.

What are Deepfakes?

• About:
  • Deepfakes are synthetic media that use Artificial Intelligence (AI) to manipulate or generate visual and audio content, usually with the intention of deceiving or misleading someone.
  • Deep fakes first came into notice in 2017 when a Reddit user posted explicit videos of celebrities. After that several instances have been reported.
• Deepfake Creation:
  • Deepfakes are created using a technique called generative adversarial networks (GANs), which involve two competing neural networks: a generator and a discriminator.
    • The generator tries to create fake images or videos that look realistic, while the discriminator tries to distinguish between the real and the fake ones.
    • The generator learns from the feedback of the discriminator and improves its output until it can fool the discriminator.
  • Deepfakes require a large amount of data, such as photos or videos, of the source and the target person, which are often collected from the internet or social media without their consent or knowledge.
  • Deepfakes are a part of deep synthesis, which uses technologies, including deep learning and augmented reality, to generate text, images, audio and video to create virtual scenes.

What are the Uses of Deepfake Technology?

• Film Dubbing:
  • Deepfake technology can be used to create realistic lip-syncing for actors who speak different languages, making the film more accessible and immersive for global audiences.
    • For example, a video was created to launch a petition to end malaria, where celebrities like David Beckham, Hugh Jackman, and Bill Gates spoke in different languages using deepfake technology.
• Education:
  • Deep learning technology has enabled positive advancements, such as restoring lost voices and recreating historical figures to deliver engaging lessons.
    • For example, a deepfake video of Abraham Lincoln giving his Gettysburg Address could be used to teach students about the American Civil War.
• Art:
  • Deepfake technology can be used as a creative tool for artists to express themselves, experiment with different styles, or collaborate with other artists.
    • For example, a deepfake video of Salvador Dali was created to promote his museum in Florida, where he interacted with visitors and commented on his artworks.
• Amplification of the Message and its Reach:
  • Deepfake technology can help amplify the voice and impact of people who have important messages to share, especially those who face discrimination, censorship, or violence.
    • For example, a deepfake video of a journalist who was killed by the Saudi government was created to deliver his final message and call for justice.
• Digital Reconstruction and Public Safety:
  • Deepfake technology can help reconstruct missing or damaged digital data, such as restoring old photos or videos, or enhancing low-quality footage.
  • It can also help improve public safety by creating realistic training materials for emergency responders, law enforcement, or military personnel.
    • For example, a deepfake video of a school shooting was created to train teachers on how to react in such a situation.
• Innovation:
  • Deepfake technology can spur innovation in various fields and industries, such as entertainment, gaming, or marketing. It can enable new forms of storytelling, interaction, diagnosis, or persuasion.
    • For example, a deepfake video of Mark Zuckerberg was created to demonstrate the potential of synthetic media and its implications for society.

What are the Present Concerns With the Deepfakes?

• Procedural Concerns:
  • Deepfakes are a problem because they can be used for various malicious purposes:
    • Spreading propaganda, and fake news;
    • Influencing elections and public opinion;
    • Blackmailing and extortion individuals or organizations;
    • Damaging the reputation and credibility of celebrities, politicians, activists, and journalists; and
    • Creating non-consensual pornography and revenge porn.
    • Deepfakes can cause various harms, such as eroding trust in institutions, media, and democracy, and undermining the rule of law and human rights.
    • Deepfake technology can violate the privacy, dignity, and reputation of individuals, and harm the mental health and well-being of the victims, especially women, who are often the targets of such malicious manipulation.
• Legal Cocerns:
  • In the US, deep fake legality is complex. Victims can claim defamation, but removing content may be seen as censorship, potentially violating the First Amendment, which protects freedom of religion, expression, assembly, and petitioning.
  • However, the Right to be Forgotten, allows a user to request companies such as Facebook and Google, that have collected his/her data to take it down.
• Ethical Concerns:
  • Deepfakes have been employed for malicious purposes, including revenge porn and hacking facial recognition systems.
  • They undermine trust in the media and blur the lines between fact and fiction.
  • Misinformation propagated by deepfakes can be mistaken as true, leading to potential social unrest.

What are the Global Approaches to Combat Deepfakes?

• India:
  • India does not have specific laws or regulations that ban or regulate the use of deepfake technology.
  • India has called for a global framework on the expansion of “ethical” AI tools.
  • Existing laws such as Sections 67 and 67A of the Information Technology Act, 2000 have provisions that may be applied to certain aspects of deep fakes, such as defamation and publishing explicit material.
  • Section 500 of the Indian Penal Code (1860) provides punishment for defamation.
  • The Digital Personal Data Protection Act, 2023 provides some protection against the misuse of personal data.
  • The Information Technology Rules, 2021, mandate the removal of content impersonating others and artificially morphed images within 36 hours.
  • India needs to develop a comprehensive legal framework specifically targeting deepfakes, considering the potential implications for privacy, social stability, national security, and democracy.
• Global:
  • The recent world’s first ever AI Safety Summit 2023 involving 28 major countries, including the US, China, and India, agreed on the need for global action to address AI's potential risks.
    • The ‘Bletchley Park Declaration’ at the summit acknowledged the risks of intentional misuse and the loss of control over AI technologies.
  • The Global Partnership on Artificial Intelligence (GPAI) summit was held in New Delhi in December 2023. The summit concluded with the adoption of the New Delhi Declaration on artificial intelligence.
    • The declaration built consensus among GPAI members on advancing safe, secure, and trustworthy AI and commitment to supporting the sustainability of GPAI projects.
• European Union:
  • The European Union's Code of Practice on Disinformation requires tech companies to counter deep fakes and fake accounts within six months of signing up to the Code.
    • If found non-compliant, tech companies can face fines up to 6% of their annual global turnover.
  • The European Union has passed the world’s first comprehensive laws to regulate the use of artificial intelligence. The Artificial Intelligence Act (AI Act) aims to introduce a common regulatory and legal framework for artificial intelligence.
    • The draft regulation aims to ensure that AI systems placed on the European market and used in the EU are safe and respect fundamental rights and EU values.
• United States:
  • The U.S. introduced the bipartisan Deepfake Task Force Act to assist the Department of Homeland Security in countering deepfake technology.
• China:
  • China introduced comprehensive regulation on deep synthesis, effective from 2023.
  • Aimed at curbing disinformation, the regulation requires clear labelling and traceability of deep synthesis content.
  • The Regulations impose obligations on the providers and users of so-called “deep synthesis technology”.
• Tech Companies:
  • Big tech companies like Meta and Google have announced measures to address the issue of deep fake content.
    • However, there are still vulnerabilities in their systems that allow the dissemination of such content.
  • Google has introduced tools for identifying synthetic content, including watermarking and metadata.
  • Watermarking embeds information directly into content, making it resistant to editing, while metadata provides additional context to original files.

What Should be Done to Address the Menace of Deepfakes?

• Learning from Other Countries:
  • The life cycle of deepfakes can be divided into three parts – creation, dissemination and detection. AI regulation can be used to mitigate the creation of unlawful or non-consensual deepfakes.
    • One of the ways in which countries such as China are approaching such regulation is to require providers of deepfake technologies to obtain consent of those in their videos, verify the identities of users, and offer recourse to them.
    • The Canadian approach to prevent harm from deepfakes includes mass public awareness campaigns and possible legislation that would make creating and distributing deepfakes with malicious intent illegal.
• Adding Watermarks to all AI-generated Videos:
  • Adding watermarks to AI-generated videos is essential for effective detection and attribution. Watermarks reveal the content's origin and ownership, serving various purposes. They aid in attribution by clarifying the content's creator or source, especially when shared in different contexts.
    • Visible watermarks also act as a deterrent against unauthorized use, making it clear that the content can be traced back to its source.
    • Furthermore, watermarks support accountability by providing evidence of the original creator's rights, simplifying the enforcement of copyright and intellectual property protections for AI-generated content.
• Deterring Users to Upload Inappropriate Content:
  • Online platforms should take steps to educate and inform users about their content policies, and perhaps implement measures to deter the upload of inappropriate content.
• Developing and Improving Deepfake Detection Technologies:
  • This can involve using more sophisticated algorithms, as well as developing new methods that can identify deepfakes based on their context, metadata, or other factors.
• Strengthening Digital Governance and Legislation:
  • This can involve creating clear and consistent laws and policies that define and prohibit the malicious use of deepfakes, as well as providing effective remedies and sanctions for the victims and perpetrators of digital harm.
• Enhancing Media Literacy and Awareness:
  • This can involve educating the public and the media about the existence and potential impact of deepfakes. Providing users with skills and tools to verify and report suspicious content, like
    • Look for visual and audio inconsistencies in the media.
    • Use reverse image search to find the original source or similar images.
    • Use AI-based tools to analyze the quality, consistency, and authenticity of the images or videos.
    • Using digital watermarking or blockchain to verify the source and integrity of the media.
    • Educate oneself and others about deepfake technology and its implications.
• Promoting Ethical and Responsible Use of Deepfake Technology:
  • This can involve establishing and enforcing codes of conduct and standards for the creators and users of deepfake technology, as well as encouraging its positive and beneficial applications.
• Blockchain-based Deepfake Verification:
  • Use blockchain technology to create an unchangeable record of who created a piece of digital media and ensure transparency in verifying its authenticity.
  • This decentralized approach allows individuals to trace the origin and modification history of media, discouraging the creation and dissemination of malicious deepfakes.