Google’s Google Earth image generator lasted about 24 hours
The tool let users make custom images grounded in Google Earth’s satellite, aerial and 3D imagery. Google described uses including professional geospatial work, but its rollback notice did not specify which generated images breached policy or name the people and organizations whose criticism prompted the decision.5
- Google announced the image-generation feature on July 30 and said on July 31 that it was rolling it back to work on stronger guardrails.5
- News organizations reported demonstrations of sensitive fabricated scenes, including a nuclear plant in Iran and a bomb crater near a Gaza hospital; the specific images have not been independently verified in the material available for this draft.1,4
- Investigative reporter Ben Huebel warned that the feature could make verification work harder, while Google said generated images were watermarked and did not appear for other users in the main Google Earth experience.1,5
- Google has not announced when or whether the feature will return, and its public rollback statement did not spell out the criteria behind the decision.2,5
Google launched Nano Banana 2 in Google Earth on July 30, 2026, allowing web users to generate custom images based on the service’s satellite, aerial and 3D imagery. In a July 31 update, Google said it was rolling the feature back while it worked on stronger guardrails, after seeing screenshots of generated imagery that appeared to violate its policies.5
The New York Times described the interval between launch and disabling as roughly 24 hours. Google’s public explanation did not name the specific images that triggered the rollback.3,5
A generative tool inside a mapping service
Google Earth Product Manager Bryan Horowitz announced on July 30, 2026, that Nano Banana 2 image generation was available globally to Google Earth web users. Google said the tool could generate custom images grounded in the service’s satellite, aerial and 3D imagery.5
Geo News described the feature as accepting text prompts to create images from Google Earth imagery, and reported that Google had promoted uses such as education, urban planning and real-estate presentations. The Guardian reported that Google’s launch announcement also suggested using it to help students visualize the past, create real-estate plans or give a familiar place a makeover.2,4
The New York Times reported that users noted the feature lacked safeguards to prevent potentially harmful content. That report describes users’ assessment; Google’s public launch and rollback statements do not detail the safeguards present during the test.3,5
What users made—and what can be verified
The New York Post, citing 404 Media’s account of a since-deleted blog post by OSINT researcher Henk van Ess, reported that he generated images of refugees near the Mexican border, a nuclear plant in Iran, a fatal crash in Amsterdam and a hospital with a bomb crater in Gaza; the Post said he reported that none of his prompts was refused. Because the underlying post was deleted and is not independently archived in the material reviewed here, those examples remain attributed demonstrations rather than independently verified images.1
The Guardian separately reported that van Ess posted images on X depicting refugees near the Mexican border, a nuclear plant in Iran and a bomb crater near a fictional hospital in Gaza. The New York Times reported that researchers, journalists and others demonstrated images of nuclear power plants, refugee camps along borders, burned forests, flooded cities and a fatal car crash.3,4
The Guardian said its reporters tested the feature by generating an image prompted as “refugees swarming New York”; the outlet described the result as not ultra-realistic but still troubling. That test is a separate example from the images attributed to van Ess.4
Rest of World also described generated imagery of a nuclear plant in Iran, a bomb crater near a Gaza hospital and floods near the U.S. Capitol. Its account attributes these examples to user demonstrations; the specific images are not independently verified by primary evidence in this draft’s materials.6
The rollback and Google’s explanation
Google’s July 31 update said it was rolling back Nano Banana in Earth “while we work on implementing stronger guardrails.” The company said it had seen people sharing screenshots of generated imagery that appeared to violate its policies, while also noting geospatial professionals had used the feature for useful purposes.5
The New York Post reported that Google did not specify which images violated its policies. Google’s statement, as documented in the blog update, did not name particular experts, organizations or media outlets whose criticism prompted the rollback.1,5
Google said generated images did not appear for other users in the main Google Earth experience and carried an AI-generated watermark. The New York Post also reported Google’s statement that images made with Nano Banana carried a SynthID digital watermark and that the company took misinformation seriously.1,5
Geo News reported that Google had not announced when or whether the feature would return. Google’s public blog update said it was working on stronger guardrails but did not describe the proposed changes.2,5
Warnings about verification and misuse
German investigative reporter Ben Huebel criticized the feature in a LinkedIn post, calling it “absolutely irresponsible and plain bad for investigations” and warning that it would make verification work harder.1
The Guardian’s account of van Ess’s posts quoted his question, “What on earth is Google doing?” The Washington Post visual forensics reporter Evan Hill, as quoted by Geo News, described the potential for misuse as “literally boundless.”2,4
Rest of World placed the launch in the context of reported fabricated footage of a blast at Iran’s Evin prison, which it said circulated on social media within minutes of a strike and was shared by outlets including Sky News and the BBC without a watermark. Rest of World reported that Citizen Lab later assessed the footage as likely part of an Israeli-government AI-enabled influence operation and that forensic experts determined it was AI-generated within days; Rest of World did not name a forensic report or official source for those assessments.6
Rest of World also reported that the U.S. government ordered satellite companies to restrict journalists’ access to imagery in the region, citing wartime security. Rest of World did not identify the agency or official behind the reported order.6
Rest of World argued that no benefit of adding generative AI to Google Earth outweighed the harm and doubt it created, and that safeguards should be tested in wars and crises before release rather than added after public criticism. That is Rest of World’s argument, not a finding about Google’s product or the effectiveness of any future safeguards.6
Google Earth Product Manager Bryan Horowitz announced that Nano Banana 2 image generation was available globally to Google Earth web users.5
Rest of World reported that users generated examples including a nuclear plant in Iran, a bomb crater near a Gaza hospital and floods near the U.S. Capitol.6
Google updated its blog to say it was rolling back the feature while working on stronger guardrails, after seeing screenshots of generated imagery that appeared to violate its policies.5
The New York Times reported that Google disabled the feature roughly a day after it launched.3
Rest of World reported that EU AI Act transparency obligations began on this date, while describing implementation as shaky and the relevant code as voluntary.6
- Independent verification or archiving of the specific fake images generated (e.g., Iran nuclear plant, Gaza crater, U.S. Capitol floods).
- The specific U.S. government agency or official behind the reported order restricting satellite imagery access to journalists.
- The named forensic lab or official source for the assessment that the Evin prison footage was an AI-generated influence operation linked to Israel.
- Google's internal criteria for pulling the feature and any documents detailing the decision to limit the live test to ~24 hours.
- A definitive list of the experts, organizations, or media outlets whose public outcry directly prompted Google's removal decision.
- The current state and effectiveness of the 'stronger guardrails' Google stated it is working on.
- 1.
- 2.
- 3.
- 4.Original reportingThe Guardian — Google Earth’s potential disinformation nightmare
- 5.Primary documentblog — Transform any place with Nano Banana in Google Earth
- 6.
- Revision 2 · October 23, 2026 · material update
- Revision 1 · October 18, 2026