The term “Deepfake,” a blend of “deep,” representing the AI deep learning technology it’s born from, and “fake,” signaling its fabricated nature, gained first public attention in late 2017 courtesy of its namesake Reddit user. Initially infamous for its role in the creation of non-consensual pornography, deepfakes swiftly grabbed the attention of experts, policymakers, and the general public, profoundly influencing attitudes toward the authenticity and reliability of digital content. While initially referring to AI-generated videos with facial swaps, the term now encompasses a broad spectrum of synthetic media applications, spanning audio deepfakes (voice cloning), image deepfakes (photo-realistic manipulation), and text deepfakes (AI-generated authorship mimicking). As deepfake technology evolves and proliferates, so too does its definition, becoming increasingly nuanced across various disciplines and contexts.
The Oxford English Dictionary officially recognized “deepfake” in March 2023, offering a definition that aligns closely with its initial conception, emphasizing video manipulation:
“any of various media, esp. a video that has been digitally manipulated to replace one person’s likeness convincingly with that of another, often used maliciously to show someone doing something that he or she did not do.”However, within the research community, the term has expanded to encompass a broader array of AI-generated content, including images, audio, and text, each producing hyper-realistic results. While computer scientists delve into the technical aspects of deepfake generation and detection, scholars from diverse fields offer varied perspectives. Psychologists explore the cognitive and emotional impacts of deepfakes, while journalists and media scholars investigate societal implications and the challenges they pose to journalistic integrity and trust.
Government agencies have also formulated their definitions of “deepfake” to inform regulatory efforts. The U.S. Government Accountability Office (GAO) defines deepfakes as “videos, audio, or images that seem real but have been manipulated with AI,” reflecting a focus on AI manipulation. Similarly, the National Security Agency extends the definition to include any multimedia manipulated or synthetically created using AI technology, acknowledging the potential for entirely fabricated content.
Legislative measures have begun to emerge globally to address the risks associated with deepfake misuse and to define the phenomenon. The European Parliament’s groundbreaking “AI Act” mandates the labeling of artificial or manipulated media content, while some U.S. states have introduced laws targeting deepfake pornography.
For journalists, combating deepfakes presents a unique challenge, threatening journalistic integrity and credibility. However, innovative tools and guidelines are being developed to detect and debunk synthetic media, preserving the integrity of news reporting. The journalism community’s approach to defining the term aligns more closely with the AI academics, where the term evolved over the years to encompass various mediums, with the focus being on the crimes and societal ramifications. Despite concerns, deepfake technology also offers positive applications, such as its use in the HBO documentary “Welcome to Chechnya,” where it safeguarded the identities of vulnerable individuals while amplifying their stories. This duality underscores the complex nature of deepfakes, requiring a delicate balance between mitigating harms and leveraging their potential for innovation and creativity in areas like entertainment, education, and virtual communication.
In conclusion, the evolving definitions of “deepfake” reflect the multifaceted challenges posed by this technology. As deepfake creation becomes more accessible, the need for robust detection mechanisms, regulatory frameworks, and media literacy initiatives becomes increasingly urgent. However, amidst the risks and uncertainties, there is also potential for innovation and creativity in harnessing synthetic media for positive applications, such as entertainment, education, and virtual communication. Achieving this balance demands ongoing dialogue, interdisciplinary collaboration, and a shared understanding of what constitutes a “deepfake” in our rapidly evolving digital landscape.