Google removed its AI image-generation feature from Google Earth on July 31, 2026 [2], less than 24 hours after the tool's debut [2].
The rapid reversal highlights the tension between generative AI innovation and the integrity of geospatial data. Because investigators and journalists rely on satellite imagery to verify news and document human rights atrocities, the ability to create convincing fake scenes poses a significant risk to global information security.
The tool, known as "Nano Banana" [1], was launched on July 30, 2024 [1]. It allowed users to generate images within the Google Earth platform, which provides global satellite-imagery services [3]. However, the feature was pulled almost immediately after its release.
Researchers warned that the tool could be used to create deepfakes of satellite scenes [4]. These synthetic images could potentially be used to spread disinformation or mislead investigators who use the platform for verification purposes [4].
Google said it removed the feature due to these deepfake and disinformation concerns [2]. The company did not provide further details on whether a modified version of the tool would return to the platform.
This incident follows a broader pattern of AI companies struggling to balance creative capabilities with safety guardrails. The speed of the rollback—occurring within a single day [2]—suggests the potential for misuse was deemed an immediate threat to the platform's utility as a reliable source of geographic truth.
“Google removed its AI image-generation feature from Google Earth on July 31, 2024”
The removal of 'Nano Banana' underscores the critical importance of 'ground truth' in the era of generative AI. While AI image generation is useful for art or conceptual design, applying it to a platform used for forensic verification—such as documenting war crimes or natural disasters—creates a vulnerability where synthetic data could be mistaken for evidence. This move signals that for high-stakes geospatial data, accuracy and authenticity must take precedence over feature expansion.
