Hundreds of Mundane AI Apps Are Hiding Sexually Abusive Deepfake Features (www.wired.com)

A recent investigation by WIRED uncovered a disturbing network of over 400 seemingly innocuous AI apps, including a "pet portrait" generator, that secretly enable the creation of sexually abusive deepfake images and videos. These apps often masquerade as everyday services, such as language learning and fitness coaching, but can load explicit content when specific conditions are met. This scheme has successfully evaded detection by major platforms like Apple and Google, with many of these apps achieving hundreds of thousands of downloads before being removed following the investigation. The significance of this discovery raises urgent concerns for the AI/ML community regarding the misuse of technology and the effectiveness of content moderation policies. The report highlights how sophisticated tactics, such as deploying benign app exteriors and leveraging ad networks to target users, have allowed these products to slip through gatekeepers. Moreover, the involvement of Chinese developers and the deliberate obfuscation tactics employed to hide their identities point to a systemic issue that necessitates stronger regulation and oversight to prevent the emergence of similar malicious applications in the future.