The latest
Google removed its new Earth Image Generation feature on Friday, one day after launching it, following warnings that users could create realistic false scenes tied to real locations. The company said it would introduce “stronger guardrails” after an open-source researcher reported generating depictions of sensitive events without refusals. Google said generated material was watermarked and did not appear to other users in the main Google Earth experience.
Details
- Intended uses: Google Earth presented the feature as a way to visualise future projects, including how a house might look before construction, or to imagine places in earlier periods. The company demonstrated the tool with an AI-generated depiction of Pompeii before the disaster that destroyed the city. The report did not describe that example as a scientific reconstruction or historical record.
- Researcher’s test: An open-source researcher told NPR that the tool readily created false images with potential disinformation uses. His prompts produced refugees at the Mexican border, a nuclear facility in Iran, an Amsterdam crash and a Gaza hospital beside a bomb crater. “Nothing was refused,” he said. The supplied reporting does not identify the researcher or detail his testing methods.
- Verification risk: The objections centred on Google Earth’s established use by journalists and open-source investigators to check claims, match landmarks and authenticate reporting. Combining synthetic scenes with genuine satellite imagery could give fabricated material a persuasive geographical setting. The report does not establish that any images made with the feature were circulated as authentic evidence.
- Google’s assurances: Google stressed that generated images were not displayed publicly to other users through the main Google Earth experience. It also said the output carried a watermark identifying it as AI-generated. The company did not specify whether those markings remained visible after downloading, cropping or reposting, according to the supplied material.
- Open questions: The report provides no figures for the number of users who accessed the feature during its single day online or the volume of images they created. Google also did not announce a relaunch date or explain which requests future safeguards would block, leaving the scope of its planned restrictions unclear.
Background
Google’s reversal followed a similar withdrawal by Meta. In the previous month, Meta removed its Muse Image generation feature less than a week after launch, following complaints that all public Instagram accounts had been enrolled automatically. The report gives no further details about Meta’s planned changes or whether that feature will return.
What’s next
The next concrete indicator will be Google’s decision on whether to relaunch Earth Image Generation and its publication of the promised stronger guardrails. Any renewed release will show whether prompts involving conflicts, disasters, borders and sensitive facilities are restricted, and how Google labels or contains the resulting synthetic imagery.
