Google and Meta Pull AI Image Features After Public Abuse Demonstrations

In late July, Google introduced a capability on its web version of Google Earth that let users request AI-generated images for any chosen location, producing photorealistic scenes built on authentic satellite and 3D data. Researchers quickly exploited the tool, creating false visuals of refugees on the Mexican border, a fictitious Iranian nuclear facility, and a fabricated bomb crater at a Gaza hospital, leading Google to disable the feature the next day while it works on tighter guardrails.

Three weeks earlier, Meta released a feature alongside its Muse model that allowed prompts to cite public Instagram accounts as reference material for image generation, without notifying the account owners; default inclusion of adult accounts and concerns about consent sparked a three-day controversy before Meta withdrew the option, calling it a miss. Similar incidents have occurred with xAI’s child-sexualized image generator and with AI models that fabricate receipts, each time resulting in restriction after public outcry.

Forensic specialist Martino Jerian of Amped Software questioned how tools with obvious deception potential pass internal reviews, arguing that foreseeable misuse should be evaluated before launch. The episode underscores that while the specific functionalities may be retracted, the broader ability to edit geospatial imagery remains, and the legal system will now have to grapple with easily produced, location-grounded synthetic evidence.