The use of generative artificial intelligence (AI) tools to create deepfakes and manipulate images is increasingly impacting public figures, including U.S. Olympic athletes, even as they compete on the global stage. During the recent Milan games, athletes faced the proliferation of AI-generated content used for harassment and political messaging, highlighting a growing challenge in distinguishing reality from manufactured media.
Users on platforms like 4chan have generated and disseminated sexually explicit imagery of female U.S. athletes, including figure skaters Alysa Liu and Amber Glenn, and skier Eileen Gu. This phenomenon, tracked by research firms like Graphika, underscores the ease with which AI technology is being weaponized against individuals in the public eye, significantly amplifying existing online harms.
The Growing Threat of AI-Generated Abuse
Generative AI technology, particularly open-source models, has dramatically lowered the barrier to entry for creating realistic fake images and videos. These accessible tools allow individuals, even those with limited technical expertise, to fine-tune AI models for specific purposes, such as generating non-consensual sexual imagery. Cristina López G., a senior analyst at Graphika, stated that these communities have effectively adapted and optimized existing AI for their illicit goals.
How AI Enhances Image Manipulation
The proliferation of customizable AI models, often shared online as Low Rank Adaptations (LoRA), has accelerated the creation and distribution of manipulated content. These LoRA files allow users to quickly adapt open-source AI models for specific outputs, akin to how video game enthusiasts share modifications. This has led to a significant increase in the quality and realism of deepfakes, making them harder to detect.
While deepfakes have existed for years, the past 18 months have seen a rapid advancement in generative AI capabilities. The widespread availability of open-source AI models means users can download, customize, and share versions fine-tuned for generating non-consensual content. This decentralized approach further complicates efforts to control the spread of harmful AI-generated media.
Despite content moderation and automatic post deletion on platforms like 4chan, manipulated images can still escape their original environment and spread virally across the internet. Research indicates that apps like Telegram and X have become key conduits for disseminating such content, demonstrating the difficulty of containing AI-generated abuse once it is created.
AI in Political Discourse and Public Perception
Beyond personal harassment, AI tools are also being employed in political messaging, blurring the lines between genuine and fabricated content. An AI-generated video shared by the White House’s TikTok account depicted U.S. hockey player Brady Tkachuk making disparaging remarks about Canadians. Though labeled as AI-generated, the video garnered millions of views and sparked controversy.
Tkachuk expressed his displeasure with the video, noting that it did not accurately represent his voice or actions. This incident highlights the growing trend of political campaigns leveraging AI to alter or manipulate public imagery for messaging purposes, sometimes without clear disclosure.
Earlier this year, the White House shared an AI-manipulated photo on X that appeared to show a protester in distress, an emotion absent in the original image. Such instances demonstrate a pattern of using AI-generated or altered media in political communication, raising concerns about transparency and the potential for public deception.
White House officials have defended the use of these AI outputs as mere jokes. However, López G. suggests that the problem extends beyond technological harm, pointing to a cultural disconnect where creators of deepfakes may not fully acknowledge the real-world impact on individuals. This disconnect suggests that addressing the misuse of AI will require a societal shift in understanding and confronting the consequences of fabricated media.
The ongoing evolution of generative AI and its widespread application suggest that challenges related to misinformation, harassment, and political manipulation will persist. The next steps will likely involve continued efforts by researchers to track and understand these evolving tactics, while policymakers and platform providers grapple with developing effective countermeasures. The extent to which these fabricated realities impact public trust and individual well-being remains a critical area to monitor.