Google’s AI May Soon Make Phone Calls So You Don’t Have To
Today’s AI news has a clear theme: AI is moving off your screen and into the physical world — and into conversations you’d rather not have yourself. From a phone-calling assistant to a wearable charm, companies are racing to make AI feel less like software and more like something you carry with you.
Google Is Testing an AI That Calls Businesses on Your Behalf
Google is reportedly testing a feature that lets its Gemini AI make actual phone calls to businesses for you, according to TechCrunch. The initial rollout is limited to U.S. users of the Pixel 11 phone who pay for a Gemini subscription. Scheduling a dentist appointment, confirming a restaurant reservation, asking about store hours — Gemini would handle all of it, speaking to whoever picks up on the other end.
Think of it like having a personal assistant who doesn’t mind being put on hold. You tell Gemini what you need, it calls the business, navigates the conversation, and reports back with the result. This isn’t a pre-recorded bot reading from a script. The AI generates responses in real time based on what the person on the other end says.
For people who dread phone calls — whether from anxiety, a busy schedule, or a language barrier — this could genuinely change how they interact with local businesses. You’d never have to sit on hold again. The catch right now is the limited availability: Pixel 11 only, U.S. only, subscription required.
Why this matters: Phone calls are one of the last friction points in daily life that technology hasn’t smoothed out. If this works reliably, it’s a meaningful shift in what people expect AI to actually do for them.
“Make phone calls to businesses on your behalf, handling tasks like scheduling appointments”
Oracle Is Protecting Itself If Its Massive AI Data Center Runs Late
A force majeure notice — a legal declaration that circumstances beyond a company’s control may prevent it from meeting a contract — is not normally the kind of document that makes tech news. But Oracle reportedly sending one for its New Mexico Stargate data center is worth paying attention to, according to TechCrunch.
The Stargate project is part of a broader push to build the infrastructure that powers large AI models. Data centers house the thousands of computers needed to train and run these systems. Oracle’s New Mexico site is supposed to be finished by 2028. The force majeure notice essentially tells whoever Oracle made commitments to: if we’re late, we may not owe you penalties for it.
For everyday people, this is a signal about how strained AI infrastructure development has become. Demand for data center capacity is enormous right now, and building these facilities is harder than it sounds — constrained by permitting, power grid access, and supply chains for specialized hardware. When a company as large as Oracle starts filing legal protection against delay, it suggests the timeline for expanding AI infrastructure is less certain than the industry likes to advertise.
Why this matters: Every AI product you use depends on data centers being built on time. Delays here ripple outward to everything from cloud storage to the AI assistants people use daily.
“Oracle may not have to pay penalties if data center isn’t ready by 2028”
Meta Made an AI You Can Clip to Your Bag
Meta has reportedly designed a small AI-powered device called the Muse Charm, according to TechCrunch. It’s shaped like a charm — think keychain-sized — and attaches to bags or clothing. The design deliberately echoes the Tamagotchi, the pocket-sized digital pet that was everywhere in the late 1990s. The goal appears to be making AI feel like a fashion accessory rather than a gadget.
The Tamagotchi comparison is deliberate and smart. That device succeeded because it made technology feel personal and cute rather than intimidating. Meta seems to be betting that younger consumers will adopt AI more readily if it looks like something they’d want to wear rather than something they’d have to learn.
What the Muse Charm actually does day-to-day isn’t fully detailed yet. But the product fits into a growing category of ambient AI devices — hardware designed to blend into your life rather than demand your full attention. It’s a different bet than the one most tech companies are making, which is packing AI into phones and laptops people already own.
Why this matters: If AI is going to reach people who aren’t interested in tech for its own sake, it may need to arrive as something that doesn’t look like tech at all.
Also Happening in AI
On the developer tools side, several quiet but useful updates landed this week. OpenAI pushed version 3.19.1 of its Python library — Python libraries are packages of code that let developers build applications on top of AI models — fixing a bug with how certain data inputs were handled. LangChain, a popular framework for building AI-powered apps, released Core version 1.6.5 with a fix that shortens unwieldy internal labels that had been cluttering developer logs. LiteLLM, an open-source tool for managing multiple AI model APIs from one place, hit version 1.100.3 with its own round of improvements. Meanwhile, The Verge reports that Microsoft is framing its unified Copilot app — which combines a chatbot, coding assistant, and task automation — as potentially as significant to workplace productivity as Microsoft Office once was. That’s a bold comparison, and one Microsoft will have to earn.
What to Watch
The Google Gemini phone-calling feature is the one to track closely over the next few weeks. The real test isn’t whether it works in a demo — it’s whether it handles unexpected responses, accents, and awkward hold music in the real world. Watch also for whether Apple or Samsung respond with similar features, since Pixel-exclusive AI tools rarely stay exclusive for long. The broader story is that AI is shifting from answering questions to taking actions, and that transition raises real questions about transparency — does the business you’re calling know it’s talking to an AI?