Google pulls AI image generator from Earth after misinformation

Google disabled the AI image‑generation feature in Google Earth less than 48 hours after its debut, citing “people sharing screenshots of generated imagery that appear to violate our policies” [2]. The rapid rollback underscores the risk that synthetic satellite visuals pose to the credibility of a platform long trusted by analysts, journalists and the public.

Immediate trigger and response

Google’s rollout integrated its Nano Banana 2 text‑to‑image model into Google Earth, allowing users to overlay AI‑generated scenes on real satellite, aerial or 3D imagery [2]. Within hours, BBC Verify testers produced a range of fabricated images—including a non‑existent nuclear facility in Iran and a sinkhole swallowing the Great Pyramid of Giza—demonstrating how the tool could be weaponised for disinformation [2]. After the tests were publicised, Google announced a pause while it “works on implementing stronger guardrails” and noted that the generated images carried invisible watermarks, though external tools could strip those marks [2].

Broader implications for geospatial trust

Google Earth has been a cornerstone for open‑source intelligence (OSINT) because satellite imagery is traditionally viewed as highly reliable [1]. Experts warned that embedding AI‑fabricated content into such a trusted map could erode public confidence, especially in fast‑moving conflict zones where verification resources are scarce [2]. Bill Greer, a geospatial analyst, highlighted that “satellite imagery has historically been seen as an especially trustworthy source of data,” and that falsified images could further diminish trust as governments tighten access to real data [2]. The incident also raises questions about the effectiveness of AI‑generated watermarks, as BBC Verify found ways to bypass both the watermark and Google’s Gemini chatbot checks [2].

Industry reaction and the “AI slop” narrative

Commentators framed the episode as part of a wider “AI slop” problem, where generative models produce low‑value or misleading content that overwhelms platforms [3][4]. Critics argued that Google’s attempt to embed AI across products without robust safeguards reflects a broader push to monetize AI at the expense of user safety. The backlash was swift on social media, with users and observers questioning the rationale for releasing a feature that could so easily be abused [3].

What to watch

• Google’s next rollout plan – Whether the company will re‑introduce the feature with stricter content filters or abandon it altogether.
• Policy updates from geospatial platforms – Responses from competitors like Esri or open‑source mapping services that may tighten verification tools.
• Regulatory scrutiny – Potential investigations into AI‑generated misinformation in geospatial data, especially as governments consider tighter controls on satellite imagery.

The episode highlights a tension between rapid AI integration and the need to preserve the integrity of trusted data sources. As Google and other tech firms iterate on generative tools, the balance between innovation and misinformation risk will remain a critical test for the industry.