The “one-click remove clothes” feature has sparked controversy. We conducted a real test on the “Baidu AI Picture Assistant.”

The news about the “Huawei Pura70 phone being able to remove clothes with one click” has attracted attention and sparked discussions among netizens, raising concerns about security and privacy. In response to this feature, TechWeb conducted some tests on the recently launched “Baidu AI Picture Assistant.”

Currently, users can directly search for images on the PC and app versions of Baidu and edit them online.

Baidu Search supports various AI functions such as removing watermarks, restoring image quality, smudge removal, partial replacement, AI image enlargement, AI redrawing (internal testing), and AI similarity search (internal testing).

In this test, we mainly focused on the “smudge removal” and “partial replacement” functions.

First, we had ChatGPT4 generate an image of a lady wearing an evening dress.

Then, we selected the area we wanted to smudge and clicked “generate immediately.”

After using the smudge removal function, we found that the “Baidu AI Picture Assistant” could not remove the clothes but only blurred them and changed their color depth.

However, we found that if we used the partial replacement function and input words like “sexy” in the “description of modified content” field after smudging the area to be replaced, the clothes could still be removed. The resulting image looked like this:

Even if we smudged the entire image, we could change the style of the clothes, like this:

While some users are curious and appreciative of this innovation in AI image editing, believing it simplifies the complexity of image editing, others express concerns about privacy infringement, fearing that this feature could be abused and pose a threat to personal privacy.

In the rapidly developing field of AI technology, balancing innovation and privacy has become an urgent issue. Based on our tests, it is evident that large-scale AI models may have vulnerabilities when facing potentially ethically risky applications. This underscores the need for relevant companies to conduct rigorous evaluations when developing features, guiding the healthy and sustainable development of AI technology.