Google Earth's "Create images" tool launched on July 30, 2026, integrating the Nano Banana 2 image-generation technology so users could generate new imagery within seconds using Google Earth's satellite, aerial, and 3D map data, according to Liberty Times Net (LTN) (E6). The tool was taken down the very next day. Commercial Times (CTEE) reported that Google "withdrew the artificial intelligence (AI) image-generation feature it had launched the day before in Google Earth, because images shared by some users were suspected of violating company policy" (E1). Central News Agency (CNA) confirmed the same sequence, reporting that Alphabet-owned Google said it pulled the feature it had launched a day earlier for the same reason (E11).
The feature was driven by "Google Nano Banana 2," an AI model that let users type a text prompt for any location on Earth and generate a realistic-looking image based on Google Earth's satellite, aerial, and 3D imagery, according to CTEE (E2). CTEE reported that Agence France-Presse (AFP) tested the tool and found it could easily fabricate satellite images with potential global impact — including fake scenes of a Paris explosion, an Iranian nuclear facility, Russian bomb craters, and a training camp for the Islamic State (IS) militant group in Syria (E3).
LTN quoted digital-investigation expert Henk Van Ess, who said Google had spent 20 years building a system the world relies on to verify facts and check information, and warned that "now it has added a button for fabricating content out of thin air." He cautioned that images faked using the tool need not look flawless, "because they inherit the credibility of the map they are attached to" (E7).
Separately, AEI satellite-imagery analyst Brady Africk warned, per LTN, that the new feature "makes it easier to produce fake satellite images convincing enough to spread rapidly online and mislead the public," and that as such AI-generated fakes become more common, they could erode public trust in satellite imagery, making fact-checking harder for journalists and researchers (E8).
On the social platform X, Google said, per CTEE, "We have seen geospatial professionals use this feature for a variety of practical purposes, but we have also seen people share screenshots of generated images suspected of violating our policies" (E4).
According to LTN, Google stated that images produced by the feature carry an invisible SynthID digital watermark that users can verify through the Gemini chatbot, and that the company prohibits generating harmful content and does not allow the spread of false, inaccurate, or misleading information, adding: "We take misinformation seriously" (E9).
CNA reported that Google's statement also emphasized public trust in the platform itself: "We understand that the public particularly trusts Google Earth to present a credible view of the world" (E12).
Google is not the only platform to have pulled back an AI image feature this year. CTEE reported that Meta Platforms, Facebook's parent company, suspended its Muse Image feature earlier this month; that tool had let users generate images using public Instagram accounts but drew criticism from Hollywood unions and others over privacy concerns (E5).
On the watchdog side, LTN reported that two organizations that verify geolocated imagery — the UK-based non-profit Centre for Information Resilience (CIR), which investigates disinformation, and GeoConfirmed, which uses geolocation techniques to verify the authenticity of conflict imagery worldwide — both called on Google to reconsider or withdraw the new feature (E10).
| Date | Event | Source |
|---|---|---|
| 2026-07-30 | "Create images" tool launches in Google Earth, built on Nano Banana 2 | LTN (E6) |
| 2026-07-31 | Feature withdrawn, one day after launch; experts Van Ess and Africk warn of credibility risk; CIR and GeoConfirmed call for reconsideration | LTN (E6, E7, E8, E10) |
| 2026-08-01 | Google confirms the withdrawal and cites suspected policy violations; details SynthID watermarking safeguard | CTEE (E1, E4), CNA (E11, E12) |
The one-day gap between launch and withdrawal — confirmed independently by CTEE, LTN, and CNA — shows Google acting on user reports faster than it could contain them: AFP's own testing (E3) surfaced fabricated images of an explosion, a nuclear site, and a militant training camp within that same window. Google's stated safeguard, the invisible SynthID watermark verifiable through Gemini (E9), addresses detection after the fact, but Van Ess's warning that faked images "inherit the credibility" of the map itself (E7) and Africk's concern about eroding trust in satellite imagery generally (E8) point to a risk a watermark alone may not resolve for viewers who never check it. That gap between Google's technical fix and the outright withdrawal sought by CIR and GeoConfirmed (E10) — paired with Meta's separate suspension of Muse Image over unrelated privacy concerns (E5) — suggests platforms are still working out how to respond to AI image-generation risks even as Google's own statement (E12) acknowledges how much public trust rests on Google Earth's credibility.
The feature was driven by Google's Nano Banana 2 AI model, which let users type a text prompt for any location and generate a realistic image based on Google Earth's satellite, aerial, and 3D data, according to Commercial Times (CTEE).
According to Liberty Times Net (LTN), Google said images produced by the feature carry an invisible SynthID digital watermark that can be verified through the Gemini chatbot.
Yes. Commercial Times (CTEE) reported that Meta Platforms suspended its Muse Image feature earlier the same month after facing criticism from Hollywood unions and others over privacy concerns.
According to Cnyes News, Foxconn (鴻海) subsidiary HCM International Company plans to sell 4.337 million Galaxy Digital shares (1.11% stake) to realize gains on a 2018 investment. Separately, MoneyDJ reports Foxconn's Hongyang Venture Capital has already exited Yang Cheng Technology, booking a NT$32,063,809 realized gain.
According to Liberty Times and UDN (Economic Daily News), Taiwan's 13 listed financial holding companies posted combined H1 2026 after-tax profit of NT$440.875 billion, up about 101% year-on-year — a record for the period, with 11 of the 13 firms setting new H1 highs.
According to CNA (中央社), Taiwan's TAIEX plunged more than 3,700 points over three sessions before rebounding a record 3,186.45 points on July 31 to close at 43,119.75 — yet the index still ended the week down 535.09 points and down 3,006.16 points for July, while listed-company market capitalization slipped to NT$140.8481 trillion.