Wired investigated 400 apps and websites linked to Chinese-language developers that appeared as regular AI tools. These applications contained hidden features that allowed users to create violent AI-generated images and video. The malicious functions were not visible to the average user. The apps were likely designed to evade detection by authorities and safety filters.
The investigation found that the hidden features were integrated into the user interface of the applications. These tools bypassed standard safety protocols to generate harmful content. The issue raises significant concerns about content moderation and safety filters on these platforms. The apps were marketed as legitimate AI tools but contained backdoors for generating prohibited material.
This discovery highlights the difficulty of policing AI tools in the global market. Developers are finding ways to hide dangerous capabilities within benign-looking software. Users and regulators need to be aware of these risks when adopting new applications. Developers building with AI must ensure their tools do not inadvertently provide access to harmful content generation capabilities.
