This paper examines the rapid evolution and growing societal impact of deepfakes—hyperrealistic, synthetic, or manipulated audiovisual media created using advanced artificial intelligence (AI) techniques. Emerging from the intersection of computer graphics and machine learning, tools such as Generative Adversarial Networks (GANs) and GPT-based models now enable the creation and manipulation of lifelike images, audio, and video with minimal expertise. While deepfakes have positive applications in areas like entertainment, education, and marketing, they also present significant ethical and security challenges, including disinformation, propaganda, and image-based abuse.

Deepfakes are frequently exploited for political manipulation, election interference, and propaganda, posing threats to democratic processes and eroding public trust in media. They facilitate the “liar’s dividend,” which undermines confidence in authentic media, challenges legal norms, and perpetuates harm to marginalized groups, exacerbating gender-based inequalities and societal polarization. Moreover, they increasingly threaten personal privacy and security, with 98% of online deepfakes now consisting of (mostly non-consensual) sexualizing content, primarily targeting women. Additionally, deepfake technology is being leveraged for crimes such as impersonation and fraud, deceiving individuals and organizations alike.

Regulatory frameworks, such as the EU’s Digital Services Act and AI Act, address some of these issues, e.g. by mandating content labeling for certain deepfake applications and enhancing social media moderation. However, stronger regulations are required, particularly to criminalize non-consensual sexualizing deepfakes. Encouragingly, the EU and several countries from around the world have recently taken significant steps in this direction. However, enforcement of existing regulations remains inadequate, not least due to insufficient resources for law enforcement and the sheer volume of deepfake content online. Therefore, advancements in detection technologies, including forensic tools for identifying manipulated media, are critical to combating unlabeled or misleading deepfakes.

In addition to technological and regulatory measures, societal strategies are crucial for mitigating the technology’s adverse impact. Enhancing media literacy can empower the public to recognize and resist deceptive content, while providing technical and financial support to journalists and media organizations can enhance their ability to verify manipulated media effectively. Ethical education for AI developers is equally important, fostering accountability in the creation and deployment of these tools. Moreover, developers and companies offering deepfake technologies must adopt stricter policies and safeguards to prevent misuse.

Looking ahead, deepfakes are poised to play an increasingly influential role in digital interactions, with applications spanning from lifelike digital assistants to metaverse avatars. However, their potential for malicious use, coupled with significant ethical and political concerns, necessitates balanced societal, regulatory, and technical responses to maximize their benefits while effectively mitigating their risks.

The opinions expressed in this text are solely that of the author/s and do not necessarily reflect the views of the Israel Public Policy Institute (IPPI) and/or its partners.