The issue of detecting deepfakes in the Global South goes beyond just the technological capabilities of current models. One of the major challenges is the disparity in quality of media from different regions. In many parts of the world, including Africa, low-cost Chinese smartphone brands are prevalent, producing media of lower quality. This poses a significant challenge for deepfake detection models, as they were originally trained on high-quality media. The lower quality of media from these regions can result in false positives or negatives, making it difficult for detection tools to accurately identify manipulated content.
Dangers of Misidentification
It’s not just generative AI that poses a threat when it comes to manipulated media. Cheapfakes, which involve simple manipulations like editing audio or video, are also common in the Global South. These cheapfakes can often be mistakenly flagged as AI-manipulated by faulty models or untrained researchers. This misidentification can have serious repercussions on a policy level, potentially leading lawmakers to crackdown on a problem that doesn’t actually exist. The risk of inflating numbers of AI-generated content could have far-reaching consequences and hinder the development of effective solutions to combat misinformation.
Creating effective detection tools for deepfakes is not a straightforward process. It requires access to energy and data centers, resources that are often lacking in many parts of the world. Without the necessary infrastructure, researchers struggle to build, test, and run detection models. This creates a significant barrier to developing local solutions for detecting deepfakes. Researchers in regions like Ghana are left with limited options, such as using costly off-the-shelf tools, inaccurate free tools, or seeking access through academic institutions. The lack of local alternatives hinders the progress in combating manipulated media and reinforces the reliance on external resources.
Challenges of Collaboration and Verification
One of the challenges faced by researchers in the Global South is the issue of collaboration and verification. Sending data to external partners for verification can result in significant delays, with lag times of weeks before confirmation of AI-generated content. This delay poses a serious threat, as the damage from manipulated media can already be done by the time verification is completed. Additionally, partnering with external institutions for verification can limit the autonomy and control of local researchers over their data. This lack of control over the verification process further underscores the challenges faced in combating deepfakes in the Global South.
While detection of deepfakes is essential, focusing solely on this aspect may divert resources from building a more resilient information ecosystem. Investing in news outlets and civil society organizations that promote public trust is crucial in combating the spread of misinformation. Redirecting funding towards these organizations can help foster a sense of credibility and authenticity in media, ultimately creating a more robust defense against manipulated content. However, there is a concern that the current allocation of resources may not prioritize these critical components of a healthy information environment. In order to effectively address the challenges of detecting deepfakes in the Global South, a holistic approach that combines detection tools with support for trustworthy media and institutions is essential.