Tencent Launches AI-Generated Content Detection Tools for Images and Text
On January 17, Chinese tech giant Tencent introduced AI-generated image and text detection tools.
Developed by the Zhuque Lab under Tencent‘s Hunyuan Security team, the detection tools utilize advanced AI technology to determine whether an image or text has been AI-generated within seconds.
The Zhuque Lab’s detection system distinguishes between authentic and AI-generated images by analyzing their multi-dimensional differences. Key aspects include:
Logical Inconsistencies: AI-generated images often feature elements that defy common sense, such as a dog with wings or a cat smoking a cigar—scenarios clearly incompatible with reality.
Explicit or Implicit Watermarks: Many AI-generated images contain visible or hidden watermarks identifying them as synthetic. While these markings might not be apparent to the naked eye, the detection tool can recognize them with precision.
Latent Features: AI-generated images may also exhibit features invisible to the human eye. For instance, when adjusting the HSV color space, AI-generated images often show localized dense bright spots in textures, a characteristic absent in real photos.
SEE ALSO: WeChat Launches the Gift-Giving Feature
To ensure high accuracy, the system was trained on a dataset of 1.4 million positive and negative samples, covering diverse generation scenarios.
In addition to images, Zhuque Lab also introduced an AI-generated text detection tool. It analyzes extensive datasets of both human-written and AI-generated text to train models capable of identifying subtle distinctions.
Similar to the image tool, the text detection system employs comparative analysis, measuring overlap between the analyzed content and AI model outputs to estimate the likelihood of AI authorship.
Tencent revealed that Zhuque Lab initially developed these tools to optimize training datasets for its Hunyuan AI model. However, as AI-generated content proliferates, the lab aims to address growing societal and industry demands for reliable detection solutions.
Zhuque Lab also plans to expand its capabilities by developing tools for detecting AI-generated videos.
