Google于周四推出的Google Earth AI图像生成功能因传播虚假信息的风险而在一天内被撤下[1][2]。该功能采用名为Nano Banana 2的AI图像生成器，允许用户通过文本提示编辑卫星、航拍和3D地球图像，创建虚假场景[2][3]。Google产品经理Bryan Horowitz在7月30日的公司博客中宣传了这一功能[3]。
用户迅速分享了生成的虚假内容示例，包括难民营和炸弹坑等有害场景[1][2]。虽然Google最初声称已实施了数字水印和防止有害主题创建等防护措施[1]，但这些措施未能有效阻止不当使用。Google随即撤回了该功能，声称"我们看到地理空间专业人士使用此功能用于各种有用的目的，但我们也看到人们共享的生成图像似乎违反了我们的政策"[2]。Google表示将在加强防护措施的同时回滚该功能[2]。
Google launched an artificial intelligence image editing feature for Google Earth on Thursday that enabled users to generate false satellite imagery through text prompts, but withdrew the tool within a day following widespread criticism over its potential to spread misinformation.[1][2][3] The feature, powered by an AI image generator called Nano Banana 2, allowed users to create custom images by combining Google Earth's satellite, aerial, and 3D imagery with AI-generated content.[3]
The company had initially claimed to have implemented safeguards, including digital watermarking and restrictions on creating harmful content.[1] However, after the feature launched on July 30, users quickly shared examples of generated images that appeared to violate Google's policies.[2] Google acknowledged the problem in a statement, saying: "We saw geospatial professionals using this feature for a variety of useful purposes, but we also saw generated images that people shared that seemed to violate our policies."[2] The company announced it would roll back the functionality while strengthening protective measures.[2] The rapid reversal drew particular attention when a BBC journalist sarcastically noted that the new AI image generation capability "couldn't possibly be misused to spread false information."[2]
