Google Pulls AI Image Tool From Google Earth After 24 Hours

Google suspended an AI image generation feature from Google Earth less than 48 hours after its Thursday launch, citing concerns over misinformation and policy violations. The tool integrated Google's Gemini Nano Banana 2 model into the platform, allowing users to generate realistic fake images from text prompts and overlay them on real satellite, aerial, and 3D imagery. Users could download and share the generated content on other platforms.

During testing, researchers and independent users demonstrated the tool's ability to create convincing deepfakes including a collapsed Eiffel Tower, a sinkhole engulfing the Great Pyramid of Giza, Russian tanks in Kyiv, fake refugee camps on the US-Mexico border, a non-existent nuclear power plant in Iran, and a fake hospital in Gaza. Other users created images depicting floods, tornadoes, and 9/11-style attacks. AI and misinformation experts highlighted a critical vulnerability: fabricated imagery "welded to genuine coordinates, drawn on genuine imagery" inherited the credibility of the map itself, regardless of how convincing the fake image appeared in isolation.

Experts warned the feature posed particular risks in conflict zones and crisis situations where information is scarce and events move quickly. Researchers noted that journalists might verify suspicious images, but not all users would conduct such checks. Testing also revealed that some of Google's built-in safeguards could be bypassed with minor prompt modifications, and that external AI detection tools sometimes failed to identify the generated content. Google stated that all AI-generated images included invisible SynthID watermarks and could be verified through its Gemini chatbot or Lens feature, though BBC Verify found these checks were not foolproof.

In its statement on Friday, Google said it had observed "people sharing screenshots of generated imagery that appear to violate our policies" and acknowledged that "people uniquely trust Google Earth for a reliable view of the world." The company stated it was "rolling back this feature in Google Earth while we work on implementing stronger guardrails." Google noted that generated images had not appeared in the main public Google Earth experience and were watermarked as AI-generated. The decision reflected broader industry challenges with AI safety; Meta had similarly removed its Muse Image feature in July after less than a week following user backlash over privacy concerns.