Google Earth’s AI Image Tool Lasted One Day Before Things Got Dark
Today’s AI stories share an uncomfortable thread: the gap between what AI can do and what it probably should do. A mapping tool generates fake disaster zones. A voice startup raises millions to make robots sound indistinguishable from humans. And a court case drags on over who actually owns the internet’s raw material. These aren’t future problems — they’re happening right now.
Google Built a Fake Satellite Image Tool, Then Deleted It Within 24 Hours
Google added an AI image generator to Google Earth — the same app millions use to explore real places on the planet. Users could describe a scene in text, and the tool would overlay a generated image onto actual map coordinates. Within a single day, people were using it to place fake refugee camps, bomb craters, and fabricated disaster zones on top of real neighborhoods.
Think of it like giving someone a photo editor tied directly to a live news map. The underlying locations are real, the coordinates are accurate, and suddenly the fake content carries the weight of something official. That combination — real place, false image — makes the misinformation much stickier than a standalone fake photo would be.
Google pulled the feature before most users even knew it existed, as The Verge reported. The damage, though, is worth thinking about. False imagery of specific, named places can travel fast on social media. Fact-checkers can’t always move faster than a screenshot.
Why this matters: Attaching generative AI — tools that create new images, text, or audio from scratch — to authoritative platforms like Google Earth creates a shortcut for disinformation that’s harder to debunk. The tool is gone, but the question of how to prevent this happening again is very much open.
“Google removed AI image generator from Google Earth after misuse generating false images.”
A Startup Raised $13 Million to Make AI Phone Calls You Can’t Tell Are Fake
According to TechCrunch, Smallest.ai has raised $13 million to build voice AI capable of holding phone conversations that sound genuinely human. The company’s pitch is speed and realism: an AI that responds in real time, without the robotic pauses or flat intonation that usually give automated calls away.
Voice AI works by converting text into spoken audio, but the hard part is making it feel natural — the small hesitations, the rise and fall of tone, the rhythm of real conversation. Smallest.ai is reportedly focusing on the latency problem, which is the delay between what you say and when the AI responds. Cut that delay enough, and the conversation feels live.
For everyday people, this raises something worth sitting with. If you receive a phone call and the voice sounds warm, conversational, and responsive, you currently have some ability to sense when it’s a machine. That ability may soon be gone. TechCrunch’s full report covers what the company says it plans to build next.
Why this matters: Ultra-realistic voice AI doesn’t just enable better customer service bots — it also hands scammers a much more convincing tool. The funding milestone signals that this technology is moving fast, well ahead of any regulations designed to govern it.
“Smallest.ai raises $13 million for ultra-fast voice AI.”
Reddit Is Still Fighting Perplexity AI Over Who Gets to Use the Internet’s Content
Reddit is keeping its lawsuit against Perplexity AI alive in court. At the center of the case is a practice called web scraping — when a program automatically crawls websites and copies their content at scale. Reddit claims Perplexity used scrapers to collect Reddit posts and comments, bypassing the access controls Reddit had put in place, and then used that content to power its AI search tool.
Perplexity’s product works by reading across the web and summarizing information in direct answers, rather than showing you a list of links. Reddit’s argument, as Ars Technica explains, is that this crosses a legal line when it involves scraping content that Reddit has explicitly blocked. DMCA — the Digital Millennium Copyright Act — is the legal framework Reddit is using to make its case.
For the millions of people who post on Reddit, this case is about something concrete: who profits from what you write online. AI companies have built products on top of years of human conversation, and courts are still working out whether that was ever legal.
Why this matters: This lawsuit is one of several running right now that will help define whether AI companies can freely train on public internet content. The outcome could reshape how AI products are built and priced.
“Reddit suing Perplexity AI over illegal scraping for search tool training.”
Also Happening in AI
OpenAI quietly released version 2.52.0 of its Python library — a set of tools developers use to plug OpenAI’s models into their own software — adding new content provenance checks designed to verify where AI-generated content originated. Music industry heavyweights are also moving: the three major record labels have reportedly proposed banning AI-generated songs from chart eligibility, a direct challenge to AI-made music’s growing presence on streaming platforms. And on a stranger note, Ars Technica reports that one startup CEO is requiring job applicants to get tattoos as part of the interview process — a story that says something about how some corners of the AI startup world are thinking about commitment and culture. All of this lands on the same week Google Earth’s AI editing tool came and went in a single day, with Ars Technica also noting the tool could generate convincing fake satellite imagery before it was pulled.
What to Watch
The Google Earth incident and the Smallest.ai funding round are pointing at the same underlying problem: AI tools are reaching the general public faster than any disclosure or verification standards exist to govern them. Watch for whether Google publishes any explanation of what internal review process missed this — and whether voice AI companies face any regulatory pressure to label their calls as automated. The Reddit lawsuit, meanwhile, is one to follow closely; a ruling in Reddit’s favor could force AI search tools to renegotiate data access across the entire web.