One of the major challenges faced in the Global South when it comes to deepfake detection is the quality discrepancies in media production. Unlike in Western countries where high-quality media is commonly used, many parts of the world, including Africa, rely on cheaper Chinese smartphone brands that offer stripped-down features. As a result, the photos and videos produced are of much lower quality, making it difficult for detection models to accurately identify deepfakes. Background noise in audio, as well as compressing videos for social media, can also lead to false positives or negatives, further complicating the detection process.
Another issue highlighted is the inaccuracy of free, public-facing tools that are often used by journalists, fact-checkers, and civil society members. These tools are not only ineffective in dealing with the inequity of representation in the training data but also struggle to cope with the challenges posed by lower quality media. This can result in content being mistakenly flagged as AI-generated when it is not, leading to potential repercussions on a policy level. Developing new and more accurate tools is not a simple task and requires resources such as energy and data centers, which are often lacking in many parts of the world.
The absence of local alternatives further compounds the issue, as researchers and individuals in the Global South are left with limited options when it comes to deepfake detection. Access to off-the-shelf tools can be cost-prohibitive, while free tools are unreliable. Academic institutions may provide some access, but the lack of local solutions makes it challenging to run effective detection models. For instance, researchers in Ghana are partnering with a European university to verify content, highlighting the significant lag time in the verification process. By the time it is confirmed whether content is AI-generated, the damage may have already been done.
While detection of deepfakes is important, there is a concern that focusing too much on this aspect may divert funding and support away from organizations and institutions that contribute to a more resilient information ecosystem. Instead of solely investing in detection technology, funding should also be directed towards news outlets and civil society organizations that can foster public trust. However, there is skepticism about whether funding is currently allocated appropriately to address these broader issues in the information landscape.
The challenges of deepfake detection in the Global South are multifaceted and complex. From quality disparities in media production to the inaccuracy of existing tools and the lack of local solutions, researchers and individuals face significant obstacles in identifying and combatting AI-manipulated content. It is crucial for stakeholders to not only focus on improving detection technology but also to invest in building a more resilient information ecosystem that can withstand the threats posed by deepfakes.
Leave a Reply