Google pulled a new AI image-generation feature from Google Earth on July 31, 2026, one day after launching it, after users demonstrated the tool could fabricate satellite-style imagery over real locations. The feature let anyone deploy Nano Banana 2, Google's image generator, to superimpose AI-generated visuals inside the satellite mapping app used by journalists, researchers, and open-source investigators worldwide. Google framed the launch as a way to get creative with geography; within hours, critics framed it as a misinformation vector aimed at one of the last widely trusted sources of visual geospatial evidence.
The tool was prompt-based, meaning users could type a description and drop the output onto a real map location. That interaction model, combined with Google Earth's reputation as a reference source, is what drew the sharpest pushback. A BBC journalist posted sarcastically that there was no way the feature could possibly be abused to spread misinformation online, a line that circulated widely on Thursday and captured the mood among researchers who rely on satellite imagery to verify events on the ground.
Google Earth has long occupied a specific role in the information ecosystem. Reporters use it to geolocate conflict footage. Human-rights investigators use it to corroborate witness accounts. Academics use it as a baseline dataset. Layering a generative image tool directly inside that product collapsed the distinction between recorded imagery and synthetic imagery in the same interface, which is the concern critics raised almost immediately.
Key facts
- 01Google launched the Nano Banana 2 image feature inside Google Earth on Thursday, July 31, 2026, and pulled it the same day.
- 02The tool was prompt-based and let users superimpose AI-generated imagery onto real satellite maps.
- 03Google said it saw screenshots of generated imagery that appeared to violate its policies.
- 04The company said it will rebuild the feature with stronger guardrails before any relaunch.
- 05The rollback comes days after Google credited AI with fixing more Chrome bugs in June than in the prior two years combined.
In its rollback statement, Google acknowledged both sides of the reception.
“We've seen geospatial professionals using this feature for a range of useful purposes, however we've also seen people sharing screenshots of generated imagery that appear to violate our policies”— Google, company statement
The company did not say how many users accessed the feature during its brief availability, nor did it detail which specific policy violations it observed. It also did not commit to a relaunch date, saying only that the tool would return once guardrails are in place.
Google added a second line explaining the rollback.
“We're rolling back this feature in Google Earth while we work on implementing stronger guardrails”— Google, company statement
The episode lands in a week where Google has been publicly touting AI wins on the safety side of its product line. The company said it fixed more Chrome bugs in June using AI than it had in the prior two years combined, a story AI Chat Daily covered earlier this week. The Earth rollback is the counterweight: an AI feature shipped without the guardrails that Google's own security teams have spent months building elsewhere.
There is a broader point worth making. Fabricated imagery is already trivial to produce with any consumer AI image generator, Google's own included. What made the Earth integration different was not the underlying model but the surface it was embedded in. A synthetic image posted to a social network is one thing; a synthetic image generated inside the interface most people associate with authoritative satellite data is another. The friction and the framing matter.
Nano Banana 2 itself continues to ship in Google's other consumer surfaces. The rollback is specific to the Earth integration, not to the model. Google has not indicated whether other geospatial or mapping products will get similar generative features, or whether the Earth experiment is now on indefinite hold.
The pull-back also arrives against a backdrop of rising skepticism about AI product velocity. Over 1,000 AI staff signed a petition this month asking OpenAI and Anthropic to slow their release cadence. Snapchat this week stopped rewarding fully AI-generated content in its Spotlight recommendations. The pattern is companies increasingly discovering that the guardrails need to ship with the feature, not after it.
For Google, the cost of a one-day rollback is small in dollar terms and large in signal terms. Shipping and retracting a feature inside a flagship product in 24 hours suggests the internal review process did not model the misuse case that a single sarcastic tweet made obvious. Whether Nano Banana 2 returns to Google Earth in a constrained form, or disappears from the product entirely, will say more about Google's appetite for generative features in trust-sensitive surfaces than any launch blog post could.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
