Google 因安全疑慮暫停 AI 衛星影像工具
Google pauses AI satellite image tool over safety concerns
Updated at: August 3, 2026 at 03:00 AM
2026年7月下旬，Google推出了一款適用於Google Earth的人工智慧工具，該工具允許使用者透過「Nano Banana 2」模型，利用文字提示生成具備攝影質感的圖像，此舉在當時引發了軒然大波。
In late July 2026, Google made waves by launching an AI-powered tool for Google Earth that allowed users to generate photorealistic imagery from text prompts using the 'Nano Banana 2' model.
此次迅速停用，是源於專家們的警告，他們證實該工具極易創造出「地理空間深偽」影像。
The rapid suspension followed warnings from experts who proved the tool could easily create 'geospatial deepfakes.'
使用者能藉由生成捏造的軍事行動或人為災難場景，創造出逼真卻虛假的現實地理位置圖像。
By generating fabricated military movements or artificial disaster scenes, users could create realistic but false visuals of real-world locations.
批評者指出，此類圖像威脅到大眾對衛星數據的信任，而這些數據正是驗證全球事件的關鍵資源。
Critics argued that such images threaten to undermine the trust placed in satellite data, a critical resource for verifying global events.
儘管Google加入了隱形SynthID浮水印，但研究人員指出，當圖像在社群媒體上分享時，這些浮水印極易被移除。
While Google included invisible SynthID watermarks, researchers noted that these could easily be stripped away when images are shared on social media.
Google目前已停止該工具的使用，並強調在建立更強大的防護機制前，該工具將持續保持離線狀態。
Google has since halted the tool, emphasizing that it will remain offline until stronger guardrails are established.
此事件為快速人工智慧創新與企業問責制必要性之間的緊張關係，提供了一個重要的案例研究。
This incident serves as a significant case study on the tension between rapid AI innovation and the necessity for corporate accountability.
隨著生成式AI持續發展，在關鍵領域維持準確性與公眾信任，已成為全球科技公司首要關注的問題。
As generative AI continues to evolve, the challenge of maintaining accuracy and public trust in high-stakes fields remains a primary concern for tech companies worldwide.
