Google Scraps Google Earth’s AI-Generated Imagery Feature After Misinformation Warnings

Google’s newest experiment with AI inside Google Earth barely had time to seek out an viewers earlier than it was pulled. The characteristic promised a quick approach to generate satellite-style imagery for the whole lot from historic visualization to actual property mockups, however researchers and open-source investigators shortly flagged a darker upside: convincing fakes that would journey sooner than any correction. With belief in geospatial proof already underneath pressure, critics argued the device risked turning a extensively used reference map into a brand new vector for misinformation.

Google halts its newest AI experiment

When you blinked, you missed it. This week, Google quietly added a brand new AI characteristic to Google Earth, then pulled it again down in lower than 24 hours after researchers raised alarms about misinformation. The priority was easy: when a mapping product turns into a trusted reference, even a small “make it up” button can bend the web’s sense of actuality.

Google confirmed the reversal in an announcement, saying it had seen folks share screenshots of AI-generated pictures that “seem to violate our insurance policies,” and that the corporate is pausing the characteristic whereas it builds stronger safeguards. The device is not seen inside Google Earth.

What the device really did

The characteristic, labeled “create a picture”, let customers zoom right into a location and generate contemporary satellite-style imagery in seconds, utilizing Google Earth’s underlying satellite tv for pc, aerial, and 3D mapping information. Google pitched it as helpful for benign work, for instance visualizing how a panorama seemed prior to now or sketching eventualities for city improvement and actual property planning.

On paper, that appears like a neat artistic layer on prime of a well-recognized product. In observe, the output may look shut sufficient to genuine imagery that it risked being handed round as proof, particularly on fast-moving social platforms the place context will get stripped away.

Why researchers referred to as it a misinformation threat

The sharpest criticism wasn’t that AI pictures may be pretend, we already know that. It was that fabricated satellite tv for pc imagery hits a very delicate nerve: geospatial screenshots are a staple for open-source investigators, journalists, and on a regular basis fact-checkers attempting to confirm occasions remotely.

Digital investigations skilled Henk van Ess wrote that Google spent 20 years constructing a reference the world depends on, then added a button that invents issues. Brady Africk, a satellite tv for pc imagery analyst on the American Enterprise Institute, warned that convincing pretend satellite tv for pc pictures may unfold shortly and “erode public belief” in actual imagery, making verification more durable.

Assessments reported by journalists confirmed the device might be pushed towards dangerous eventualities, together with depictions tied to delicate navy or geopolitical places. That echoes current historical past: in 2023, an AI-generated picture of an “explosion” close to the Pentagon briefly rattled markets.

Google’s safeguards, and what comes subsequent

Earlier than the pause, Google stated AI-generated pictures included invisible digital markers that customers may detect with Gemini, and that it blocked era on “harmful” subjects underneath its AI use insurance policies. Critics argued that enforcement and detection are solely as sturdy because the real-world testing, and that screenshots journey farther than provenance instruments.

Google says the characteristic might return with harder guardrails. The larger query is whether or not geospatial reality, as soon as softened by generative AI, may be reliably re-hardened for the individuals who want it most.