Link: Google Earth’s AI deepfake tool only lasted one day

Google has discontinued a recently launched Google Earth feature that allowed users to overlay AI-generated images on satellite maps using text prompts. The tool, introduced on Thursday, was disabled due to concerns about the creation of misleading scenarios, like adding refugee crowds at borders or bomb craters near hospitals.

Initial responses from Google highlighted the inclusion of a digital watermark on these images and restrictions preventing generation of harmful content. However, concerns persisted about the potential misuse of the feature.

By Friday, Google acknowledged the unique trust placed in Google Earth for accurate geographical representations and noted misuse by some users. Consequently, the company decided to remove this feature to prevent the spread of potentially harmful misinformation.

Despite assurances and watermarks indicating AI creation, Digital Digging’s Henk van Ess demonstrated the feature’s potential for creating deceiving visuals. He successfully fooled advanced AI detection tools with a falsified video from Google Earth.

Van Ess’s experiments showed that the AI did not restrict any harmful prompts, suggesting a lack of effective moderation within the tool. Google has now pledged to develop stronger safeguards before potentially reintroducing this capability.

 #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.