Say no to ai generated images

The rapid integration of generative AI into satellite mapping platforms has created new vulnerabilities for visual forensic investigations. Experts warn that the ability to synthesize realistic satellite imagery undermines a historical baseline of truth, as these tools can now be used to depict events that never occurred, such as natural disasters or military strikes.

What happened

Concerns over deepfake satellite imagery gained prominence following the release of an AI-generation feature on Google Earth, which allowed users to create custom satellite-view imagery with simple text prompts. Although Google later rolled back the feature after researchers demonstrated the ease of generating disinformation, the incident highlighted the potential for abuse in the open-source intelligence community. Analysts noted that the technology could be used to create imagery of non-existent events, such as a flooded U.S. Capitol or fires at strategic oil terminals.

This risk was previously observed during the conflict involving the U.S. and Iran, where fake imagery circulated in media depicting a damaged U.S. base in Bahrain. These falsified images utilized Google Earth data combined with AI generation to simulate damage that did not reflect the reality on the ground. Researchers like Jake Godin of Bellingcat emphasize that while satellite imagery has traditionally served as a reliable verification tool, the democratization of AI-driven synthesis risks accelerating the spread of misinformation. Furthermore, the proliferation of such fakes may lead actors to dismiss genuine satellite evidence as AI-generated, creating a cycle of doubt that complicates the verification of breaking news and geopolitical events.