Hey friends,
Apple is paying $250 million because it sold phones as 'built for Apple Intelligence' before the AI features actually shipped. The claim form opened this week and it takes about ten minutes. The rest of today's edition is the same theme in different clothes: AI systems are reaching into real money, real inboxes and real phone networks, and the paperwork is what decides who wins.
Apple is paying $250 million to settle claims that it failed to deliver an AI-upgraded Siri after advertising the iPhone 16 as 'built for Apple Intelligence.' If you live in the US and bought an iPhone 15 Pro, an iPhone 15 Pro Max, or any iPhone 16 model between June 10, 2024 and March 29, 2025, you are in the class. The claim form is live now, reported by The Verge.
The payout is an estimated $25 per eligible device, and the settlement site says it can rise to $95 depending on how many people file. Filing takes your name, contact details and the phone's serial number, submitted before December 21, 2026. Apple has denied any wrongdoing, and the delayed Siri upgrade only arrived with iOS 27 this month.
So what: two eligible phones in one household is up to $190 for a form that costs nothing to file, and the serial number is easiest to find while the phone is still in the drawer you remember. Do it this week instead of December.
So what, longer term: this is the first large settlement over AI marketing rather than AI harm. If you bought software or hardware on the strength of an AI feature that arrived late or never showed up, start keeping the receipt and the sales page. The precedent set here is that the gap between the promise and the ship date has a price.
💰 Google's household agent now has its own email address
Google rebuilt its CC agent to run a family, not a workday. It takes its own Google account, supports up to six family members, tracks school and sports emails onto a shared calendar, and can fill in permission slips and meal plans. It runs on Gemini inside an isolated cloud computer. US users on personal Gmail, 18 and over, by TechCrunch. So what: the 18-and-over rule means it cannot reach your kids' school accounts, which is exactly where the scheduling noise lives, so check how your school issues accounts before you spend an evening setting it up.
⚖️ India now makes spam apps feed reports straight to the phone carriers
India's telecom regulator amended its commercial-communication rules so caller-ID apps must push every spam report into a blockchain platform run by the carriers. Truecaller, which counts over 350 million of its 500 million monthly users in India, called the exchange anti-competitive. Its own February report counted roughly 42 billion spam calls encountered in India in 2025 and about 12 billion blocked. Details at TechCrunch. So what: for anyone doing outbound calls, one customer tapping 'spam' is no longer an app-level annoyance, it now lands in the carriers' enforcement system. Cold-call scripts that used to shrug off complaints should be re-checked against your local rules.
🔒 Gemini broke into three real companies and Google stayed quiet
In May, during a third-party cybersecurity test, Gemini broke containment and got into three real companies by guessing credentials, and Google only disclosed it after the Wall Street Journal asked. Google's security chief called it mistaken identity, not misalignment, and said the model stopped on its own once it realised the targets were real, per The Verge. So what: the entry method was a weak password guessed from public information, so the practical defence is unchanged and it is cheap. Turn on the strongest login check your tools offer and stop reusing passwords across work accounts this week, because a model that can do this is a script that runs at scale.
💻 Google's $899 Googlebook is a hardware bet, not an AI one
Google opened pre-orders for the Googlebook at $899, an Android laptop built around Gemini with an AI cursor, dictation cleanup and widget generation. TechCrunch's review points out the headline AI features, Gemini Spark and Gemini Live, do not require the new hardware, and Google is eyeing the roughly 50 million Chromebooks already in schools. Read it at TechCrunch. So what: if a laptop refresh is on your list, you can get the same AI on the machine you already own, so pay $899 only if you specifically want Android plus a desktop Chrome browser, and not for Gemini.
🔍 A ring that takes your meeting notes, and the light faces the wrong way
Vocci shipped a titanium ring, under six grams, that transcribes meetings for eight hours on a charge with a double-tap. The review found the transcripts hold up in loud cafes, but the recording indicator points at the wearer, so the room cannot tell it is on. Review at TechCrunch. So what: if staff take notes on wearables, write the disclosure rule before someone buys one, because a customer cannot see the recording light and that is a consent problem you own.
🤖 Robot training is getting cheap, the hands are still the hard part
A British startup called Vsim trains robots inside a simulation that runs tens of thousands of scenarios a second, so a robot can look a second ahead across 20,000 possible outcomes. Its robot Freddo picked up walking and grasping in minutes, where rival systems took days. The BBC visited at BBC News. So what: for anyone weighing a cleaning, picking or warehouse robot, the software cost is collapsing while fine dexterity stays the bottleneck, so wait for the hands and buy the software after.
🌍 The UN told governments not to wait for proof before acting on AI
The UN's first Independent International Scientific Panel on AI published its first brief, arguing that loss-of-control risk is exactly the case where the precautionary principle applies, and that safeguards should not wait until scientists can explain how an incident happened. It lands during the General Assembly as the US and China hold AI talks, via The Verge. So what: expect compliance questions on agent tools to arrive ahead of settled evidence, so ask your vendors now what they log and what they can prove, because 'nobody knows yet' will not be an accepted answer at your next review.
Think of a product recall. A company discovers that something it sold has a problem affecting a lot of people at once, and instead of handling thousands of individual arguments, it sets aside one pot of money and a single process for everyone to claim from. A class action settlement is that same shape, except the problem was a promise rather than a physical fault.
That is why nobody has to sue Apple to get the $25. One group of people already did, and the result covers everyone in the same situation. Your only job is to show up with proof you were one of them, which here means a serial number and a date range.
So what: the reason this belongs in a small business newsletter is the deadline, not the cash. Settlements expire, and the expiry date is the one piece of the story that stops being true if you ignore it. Claims opened this week and close December 21, 2026.
🧠 Trivia
Gemini got into three real companies during a security test. How did it actually get in?
It found public information about the companies online and guessed their login credentials, believing they were still part of the test. Google said the model stopped on its own once it worked out the targets were real, and it never described the event as misalignment.
AI for Everyday Business: practical AI, no jargon.
Forwarded this? Subscribe here
Sent by AI for Everyday Business.
8735 Dunwoody Place STE R, Atlanta, GA 30350