Google launched a new artificial intelligence feature within Google Earth on Thursday that enabled users to edit satellite imagery by submitting text prompts. The tool functioned by generating modified versions of real-world locations based on user requests, effectively creating deepfake satellite imagery. However, the company deactivated the feature less than 24 hours after its initial release.
The decision followed reports that users had employed the tool to generate problematic content. A digital digging researcher demonstrated the feature’s capabilities by creating images depicting fictional scenarios, including refugees near international borders and destructive military strikes near civilian infrastructure. The researcher documented that the generated images appeared to evade content moderation systems despite Google’s stated safeguards.
In its initial response to these demonstrations, Google noted that images produced through the tool carried a digital watermark indicating AI generation and stated that the company had implemented restrictions preventing creation of content on certain sensitive topics. However, a subsequent statement released one day after launch indicated the company had identified instances where generated imagery appeared to violate its content policies, prompting the decision to discontinue the feature.
The company’s follow-up statement emphasized that Google Earth holds particular importance as a trusted source for accurate geographic information. While acknowledging that geospatial professionals had begun using the feature for legitimate applications, Google determined that the risks outweighed the benefits. The company noted that generated images did not automatically appear in the main Google Earth experience visible to other users and carried watermarks, but committed to developing enhanced content safeguards before reintroducing such functionality.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI