Three Google Gemini stories landed within days of each other this week, and together they sketch the awkward shape of the assistant's next phase: a week in which it speaks on your behalf, ships itself invisibly inside artwork, and lends its logo to people who only want your password.
Gemini May Soon Call Your Mom for You
Google's Call for Me feature arrived on the Pixel 11 in late September with a narrow, practical brief: let Gemini dial businesses, navigate automated phone menus, book appointments and check whether something is in stock so you do not have to. On October 5, Android Authority reported that an APK teardown found something far more personal hiding inside the app.
The leak points to a "Gemini Calling" introductory screen whose sample prompts read like text messages written by an extremely polite robot:
- Call Mom and tell her I will be 15 minutes late
- Call John and ask if he is coming for dinner
Taken at face value, that is an expansion from businesses to friends and family, though a modest one. The examples suggest Gemini would deliver short, message-like calls rather than hold a genuine catch-up conversation with your relatives. It could still be useful for the people who never read texts, miss calls constantly, or still keep a landline.
Android Authority also found granular per-app permissions for Gemini actions, including what look like phone call permissions. Those settings would let users opt out of receiving automated calls from Gemini at all. None of this is confirmed, and Google may never ship it. The only line the company has ever published about the feature, in a Pixel blog post, remains wonderfully vague: Gemini "might tell your mom you're running late."
Fortnite Quietly Replaced an Album Cover Bearing a Gemini Watermark
If you are going to hide an AI watermark inside album art, do not surround it with stars. Fans spotted a white, star-shaped Google Gemini watermark tucked into the bottom-right corner of the Club Ironmouse album cover, where it blended neatly into the star pattern already framing the artwork.
The record, VTuber Ironmouse's first full-length album, was released in January 2026 and sat unnoticed until Fortnite added its track "Meow" as a Jam Track on September 19. A Reddit user posted screenshots of the cover and pointed out the watermark, and commenters quickly realised the same image was also serving as the album art on Apple Music and Spotify.
Fortnite quietly swapped in a different version of the Club Ironmouse graphic, and both streaming platforms soon followed. Ironmouse, who is openly hostile to generative AI and previously severed ties with the gacha game Neverness to Everness over its use of AI assets, said she had no idea the image was synthetic.
"I had no idea thanks for bringing it to my attention," she wrote. "My team has contacted the people in charge of this particular release and demanded it to be changed."
The episode is small but instructive. Watermarks only work as provenance if somebody actually looks, and a quiet artwork swap does not undo the fact that an AI-generated image travelled through a major game and two streaming services under a human artist's name.
Scammers Are Wearing Gemini's Logo on a Fake Ad Portal
While Google was teaching Gemini to make phone calls, someone else was teaching Gemini's branding to steal them. Security researchers at Island disclosed a human-operated phishing platform that impersonates the advertising products of ChatGPT, Gemini, Claude, Perplexity, Meta Muse and Manus.
Each fake product gets its own pitch. Gemini's page promises manager account and linked client support. ChatGPT's advertises a weekly Google Ads briefing. Claude gets an entire advertising portal of its own. Clicking the all-important Connect button does not open a real Google login. Instead the page draws a browser inside the browser, with a fake address bar showing accounts.google.com or an Okta tenant while the real browser stays parked on the attacker's domain.
Behind that fake window, a live operator watches and steers. The platform records up to three password attempts, fingerprints the device, and lets the operator choose which multi-factor challenge the victim sees next, whether that is an SMS code, an authenticator code, a Google approval prompt, QR verification, number matching or an Okta push.
Island researchers Oleg Zaytsev and Ofek Ronen said the campaign targets agency staff, media buyers and manager-account administrators, with ad accounts carrying a clean spend history as the prize. Recovery is slow. "For the victim, the card is the easy part: they can remove it within hours," they wrote. "Getting the account back is not."
The freshest lure, museads.ai, appeared on September 16, roughly a week after Meta launched its Muse agent. Island's advice is blunt: verify unexpected beta programs and account connectors through official vendor sites, and inspect the real browser's outermost address bar rather than any window drawn inside the page.
Three quick takeaways from a very busy week:
- Gemini is moving from answering questions to taking actions, and those actions are getting more social.
- AI watermarks are quietly failing as provenance tools, because nobody looks until a fan does.
- The loudest Gemini-branded risk of the week was not model behaviour at all. It was a login box.
Comments