Amazon Texts to Shop, OpenAI's Cyber Model, Google and Schools

Grant: Meanwhile Google told schools, keep our AI off and you lose Search too. Lawmakers are not happy.

Claire: I'm Claire.

Grant: And I'm Grant, and this is The Week in AI, the Monday catch-up to get you ready before the work week starts, covering the week ending Sunday, November eighth.

Claire: Eight stories, let's get into it.

Grant: Claire, Amazon just turned texting into a shopping channel. The Verge, TechCrunch, The Information, Adweek and Digiday all covered it. A small group of US customers can now text Amazon's Alexa for Shopping agent, search, compare, add to cart, even buy straight from a photo.

Claire: Small group first, so this isn't everyone yet?

Grant: Right, invites go out by text to a limited set, but Amazon says millions more customers follow in the coming weeks, and if you're not invited you can join a waitlist.

Claire: And the ads are already in there? That's the part marketers will care about.

Grant: Adweek reports Amazon is testing ads inside the text replies without charging for them yet, and Digiday's headline says those replies already carry Sponsored Product ads.

Claire: So there's a free window before pricing shows up. Anything on how much this actually moves spend?

Grant: Adweek cites CEO Andy Jassy saying Alexa for Shopping users spend fifty percent more than non-users. That's Amazon's own figure, but it's the number they're leading with.

Claire: Practical takeaway for a brand selling on Amazon?

Grant: Watch how your products get chosen when someone just texts a vague request, check your listings look right in that format, and don't assume the free ad placement lasts once this rolls wider.

Claire: Grant, OpenAI launched something called GPT-6 Cyber this week, and the framing is interesting, a model built to hack things so it can also fix them. OpenAI itself, plus The Verge, TechCrunch and The Decoder all wrote it up.

Grant: Wait, OpenAI wants an AI that hacks on purpose? What stops it from wandering into systems nobody authorized?

Claire: That's the whole pitch. OpenAI rates it Critical in its own Preparedness Framework, up from High for the previous version, GPT-5.6 Cyber. But they say out-of-bounds behavior fell below one percent in testing, versus roughly five percent before.

Grant: Below one percent, that's OpenAI's own number though, right? Nobody outside has checked it?

Claire: Correct, The Decoder flags that all these benchmark figures come from OpenAI, nothing independently published yet.

Grant: So who actually gets to use this thing?

Claire: It's starting with select customers in a program called Daybreak Red, meant for authorized defensive work, with all Daybreak Red customers following in the coming weeks. Access stays gated for now.

Grant: What does a marketing team actually do with this news, since most of us aren't security researchers?

Claire: Simple. Ask your agency or vendor how fast they patch, because that clock just got shorter.

Grant: Claire, this one got heated. TechCrunch, The Verge, Ars Technica, The Information and The Decoder all covered Google's new policy for schools. Starting with student accounts, Google is switching off Gemini, NotebookLM, Gemini in Classroom, and Search, unless an administrator turns them on.

Claire: Hold on, Search is in that list? You can't just keep Search and skip the AI parts?

Grant: No, and that's the whole controversy. Google's education vice president told The New York Times the two are too intertwined to separate. If a school rejects the AI tools, students see a message that Search is unavailable.

Claire: That's a real bind for a school district. What's the timeline, and who's pushing back?

Grant: The Decoder's headline puts it starting in 2027. As for reaction, The Verge reports Senator Bernie Sanders called it extortion, New York City Mayor Zohran Mamdani said no corporation should dictate how kids are taught, and a California state official warned against strong-arming schools.

Claire: Is there anything softer in here, like a feature meant to help rather than just gate access?

Grant: Yes, Google's adding Guided Learning, a tutor-style feature built into AI Overviews, meant to walk students through a problem instead of handing them the answer. The Verge says it's being tested in select US schools this month.

Claire: So for content teams building educational material, what should they actually watch?

Grant: Watch how Guided Learning cites sources inside Search, and how many districts actually opt out versus just going along with it. If you make content for students or teachers, this AI layer is about to sit between you and them.

Claire: Grant, Anthropic gave Claude a new feature this week called Personal Finance, and it's a bigger step than it sounds. TechCrunch, The Information and The Decoder all covered it.

Grant: Claude is looking at people's actual bank accounts now?

Claire: Through Plaid, yes, in beta this week for Pro, Max and Ultra subscribers in the US. But it's read-only. Claude can't move money, invest it, or manage anything, it can only look.

Grant: Look at what exactly?

Claire: Balances, credit limits, up to two years of transactions. It'll sort spending into categories, flag duplicate charges, unexpected fees, bills that are creeping up. You can ask it to explain something like APR in plain language.

Grant: And none of that becomes ad targeting down the line?

Claire: Those conversations aren't used for model training, and the company says there are no ads and no sponsored recommendations, it doesn't even sell its own credit cards or loans.

Grant: TechCrunch mentioned ChatGPT already does something similar?

Claire: Right, OpenAI introduced comparable tools back in May, so both major chatbots now compete on this. The Decoder reports Anthropic eventually wants to connect Claude with licensed financial advisors. If you're a finance brand, that partnership layer is worth watching closely, since neutral answers are the whole selling point here.

Grant: Claire, Meta rolled out Muse Voice to everyone in the US this week. The Verge, TechCrunch and Social Media Today all have it. You can now have a real-time voice or video-style chat with Meta's Muse AI agent, and there's an animated avatar reacting as it talks.

Claire: Does it actually see the person talking, or just hear them?

Grant: Just hears, TechCrunch is clear on that. But the avatar still animates with expressions that match the tone of the conversation, and you can customize the look, plus change the voice's tone or even request an accent.

Claire: Where does this actually run right now?

Grant: Mobile, web, and Mac today. Meta's AI glasses and an upcoming device called Muse Charm are both slated to get it too, with Muse Charm expected to ship before year end.

Claire: Is there actual demand for this, or is it a feature looking for a use case?

Grant: Fair question, Social Media Today was skeptical, pointing out Meta's earlier celebrity chatbots never really caught on. Meta's own pitch leans toward things like rehearsing a pitch or a presentation out loud.

Claire: So for a content team, what's the marketing angle?

Grant: When people talk instead of type, queries get more conversational and answers get heard, not scanned. If voice spreads across glasses and Muse Charm like Meta plans, start thinking about how your brand sounds, not just how it reads.

Claire: Grant, Airbnb put a real date on its AI travel agent this week. TechCrunch and The Information both covered it, and it came inside the company's Q3 shareholder letter.

Grant: A real date, meaning more specific than before?

Claire: Yes, Airbnb had only said within the next year, now it's committing to the first half of 2027 for consumers. And it's not a chat box, CEO Brian Chesky described it as something closer to generative UI.

Grant: Meaning what, practically?

Claire: The agent builds a custom interface for each question, pulling maps, photos, comparisons, calendars and itineraries from a library of hundreds of travel-specific components, instead of just answering in text.

Grant: Does it only help you book, or does it stick around during the trip too?

Claire: It's meant to handle in-trip issues as well, reservation changes, payment problems, refunds, rebooking, even twenty-four hour safety support. Chesky argues outside agents can't easily do that kind of thing on their own.

Grant: And the business was doing fine while all this was announced?

Claire: Q3 revenue rose nineteen percent to four point eight seven billion dollars, and Airbnb raised its 2026 growth forecast to the high teens. For hosts and travel brands, the signal is that Airbnb wants discovery staying inside its own app, so it's worth checking how your listing photos and descriptions would look assembled into one of those side-by-side views.

Grant: Claire, a smaller deal but a telling one. OpenAI itself, plus The Information and The Decoder, confirmed OpenAI has agreed to buy a company called Pierre Computer Company.

Claire: What does Pierre actually make?

Grant: Their main product is Code Storage, which hosts code and tracks its history through Git. They also built tools for navigating files, viewing code changes, and live syntax highlighting, and those are already used inside OpenAI's Codex coding agent.

Claire: So this isn't a new product, it's buying something already plugged in?

Grant: Exactly, and The Decoder frames it as OpenAI making sure Git itself can keep up when people and AI agents are both editing the same codebase at once. The deal still needs regulatory approval.

Claire: Anything to reassure developers who rely on Pierre's open-source tools?

Grant: OpenAI says those open-source projects stay open source. Worth watching once the deal closes, to see if that actually holds.

Claire: What's the marketing team takeaway here, since this is pretty developer focused?

Grant: If your team uses Codex to build landing pages or automations, expect cleaner diffs and faster version history, which just makes it easier to check what the agent actually changed before anything ships.

Claire: Grant, last one, Apple. The Verge reports iOS twenty-seven point three rolled out Tuesday as a Developer Preview, and it lets developers plug Siri AI's newer features into, quote, all apps.

Grant: All apps, meaning any app on someone's phone could get full Siri AI abilities?

Claire: That's the direction. Apple had only offered a smaller set of features before. Now the assistant can use personal context, understand natural language requests, and rely on what's on screen to finish a task inside a third-party app.

Grant: There's something called app skills in here too, right?

Claire: Right, developers can bundle several tasks into one Siri action. And users won't even need to name the app if it's in the foreground, showing a Live Activity, or already part of a Siri AI conversation. Apple's own example is a ride share app where you ask Siri to change the pickup address mid-ride.

Grant: Is Apple doing this inside its own apps too, or just opening the door for everyone else?

Claire: Both, Apple's extending the same abilities to Mail, Maps, Messages, Photos, Safari and Notes, and says users can ask Siri AI what it can do inside each one.

Grant: So if a brand has an iPhone app, retail, delivery, media, what should they do this week?

Claire: Pull up the developer documentation, figure out which repeated customer actions matter most, and test how well they bundle into a single Siri skill before the wider release lands.

Grant: So if we're picking one thing out of all eight stories, Claire, what is it?

Claire: It's the same thread running through Amazon, Meta, Airbnb and Apple. Your content is about to be read, assembled, or spoken by an agent instead of displayed the way you built it.

Grant: So this week, go test it.

Claire: And on the security side, ask your vendors how fast they patch, since tools like GPT-6 Cyber mean vulnerabilities get found a lot faster now. That's the whole list, really, see how you sound and look through an agent's eyes, and make sure the plumbing behind it is patched.

Grant: That's the week. A lot of it comes down to agents doing the talking and the choosing for people now, texting, voice, visual interfaces, all of it.

Claire: If you want this kind of automation working for your own brand, Z Media builds fully automated AI content and CRM systems that write, design, film, publish and answer for you. Talk with our team, the links are in the show notes, or call or WhatsApp us at four one nine, five zero zero, four eight eight two.

Grant: And if you just want the key points every day instead of waiting for Monday, subscribe to the AI news wire. We'll see you back here next week, same time, same catch-up.

Claire: Have a good week, everybody.