AI News This Week (November 4 to 11, 2026): What Changed, in Plain Language, and What To Do About It

How we built this week's AI news roundup (and what we left out)

Most "AI news this week" pages are written for investors or developers, and a lot of them repeat rumours. This one is different on purpose. It covers what changed between November 4 and November 11, 2026, in plain language, and it only includes an item when we could open the company's own page and read it. For each item you get four things: what happened, who it affects, what you should actually do about it, and the privacy or scam angle that most news pages skip.

Here is how we sourced it. We opened Anthropic's news page and its Claude Fable 5.2 announcement, Google's blog feed and the individual Google posts linked below, the Gemini app release notes, Google's Android blog, Microsoft's official blog feed, and OpenAI's news feed. Some OpenAI pages refused to load for our tools (they returned an access error), so for those items we only use the one-line description published in OpenAI's own feed, and we say so. If a detail was not on a page we opened, it is not in this article. Where we looked for something and found nothing, for example a Microsoft consumer Copilot announcement this week, we tell you that too.

IT Cares is an independent IT support company in Quebec. We are not affiliated with Google, Anthropic, OpenAI, Microsoft or Adobe, and nothing here is sponsored. If you want the longer, evergreen versions of these topics, we link our full guides along the way, such as our free Google AI tools list and our AI privacy and safety checklist.

The week at a glance (November 4 to 11, 2026)

If you only have two minutes, this table is the whole week. Dates are the dates on the vendor's own page.

Notice what is missing. There is no emergency, no new bug you must patch, and no change that forces you to switch tools. That is normal. Most weeks in AI are a stream of new model names and small feature additions, and the useful skill is knowing which ones touch your daily life. This week, for regular people, three things matter: free video in Google Vids, the new "sketch to app" trick in Gemini, and the stream of scams that always follow big model launches.

What to do about all this: a 15-minute plan

Before the item-by-item details, here is the practical version. These steps are safe for anyone and do not require paying for anything.

1. Do nothing urgent. None of this week's announcements require action to keep your accounts safe.
2. Update the apps you already use. New features usually arrive with app updates, and updates also carry security fixes. Open your app store and update the Gemini, ChatGPT and Claude apps if you have them.
3. Check one privacy setting per tool. Open the settings of the AI app you use most and look for the chat history or "keep activity" option. Decide on purpose whether your chats are saved.
4. Use only official download pages. After a big launch, fake "download the new model" pages and copycat extensions appear. Go to the app store or the vendor's own site, never a link from a text, ad or social post.
5. Keep private data out of free chat tools. No passwords, banking details, health records or client files. See the section on privacy below.
6. Pick one new thing to try. Try one of the free features (Google Vids video or the sketch-to-app trick) with a harmless example, then stop. Trying everything is how people end up with ten half-configured accounts.

Gemini 4 Carbon: Google's most capable model, mostly for paying users and developers

What happened. On November 10, 2026, Google published "Gemini 4 Carbon is now available." Google describes Carbon as its most capable model so far, positioned above Gemini 4 Argon. In Google's words, Argon suits most coding and knowledge work, while Carbon is aimed at harder problems such as multi-step debugging and open-ended scientific research, and it uses more tokens to reason and check its work. Google says it was first released to cybersecurity defenders about two weeks earlier.

Who gets it. According to the post, developers and enterprises get paid API access through Google AI Studio and Gemini for business. Consumers need to be on Google AI Ultra: Carbon is rolling out to Ultra subscribers in Google Antigravity, the Gemini app and AI Mode in Search. Scientists can apply for six months of free access through Antigravity's highest usage tier, subject to eligibility and selection. The post does not list countries, so we cannot tell you whether it reaches Canada on day one. Check the Gemini app's model picker.

Price. For developers, Google lists an introductory price of $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after November 30, 2026. If you are a regular person, you will never see these numbers. They matter only if you pay for API usage inside an app or a tool that bills you by use.

What Google claims. The post reports leading results on several benchmarks (CursorBench 4.0, WANDR and AutomationBench) and strong results on cybersecurity tests. These are Google's own reported numbers, so treat them as the company's claims rather than independent proof. Google also reports that its safeguards blocked 99% of simulated attack trajectories, while acknowledging that "prompt injection robustness remains an area for continued improvement." That last phrase is worth remembering: even the best current model can be tricked by hidden instructions in a web page or a document. We explain this risk in our guide to AI agents and their security risks.

What you should do. If you use the free Gemini app, nothing changes today. If you pay for Google AI Ultra, look for Carbon in the model picker, but do not assume it is better for everyday chat. A stronger reasoning model is slower and heavier, and Google itself says Argon already handles most work. If you are on the free plan or the cheaper paid plans, the right answer is to keep using what you have. For a clear picture of what free users get, read our free Google AI tools list.

Scam angle. New model names are bait. Expect fake sites advertising "Gemini 4 Carbon free download," "Carbon unlocked" or "Ultra for $1." Google distributes Gemini through its own app, its own website and your phone's official store. There is no reason to install a separate "Carbon app" from a link you received.

Free AI video in Google Vids: three 15-second clips a month

What happened. On November 5, 2026, Google announced that Google Vids, its video creation and editing tool, now runs on Omni 1.2 Flash, its newest video generation model. Google lists three things: single-take product promos and social content, turning reference images into videos with characters and scenes that stay consistent from start to finish, and more natural-sounding speech narration.

Who gets it free. The post says anyone with a Google account or a Google Workspace subscription can make up to "three 15-second videos per month" at no cost, through vids.new. That is a small allowance, but for regular people it is the cleanest answer this week to the question "can I try AI video without paying?" For a deeper walk-through of Google's free video options and how credits work, see our guide on how to use Google AI video for free.

What the page did not say. The announcement does not mention watermarking, data retention or whether you can use the videos commercially. If you plan to use a clip in a business ad, read Google's terms for Vids first and keep a note of when and how you generated it. We could not confirm these points, so we will not guess.

Practical tips for your three videos.

• Plan before you spend. With only three clips a month, write a one-sentence description of each before you generate. Vague prompts waste a clip.
• Use your own images carefully. If you upload a reference photo, only use images you own and that do not show other people who have not agreed to it. Never upload a photo of a child.
• Avoid text inside the video. AI video tools still tend to garble written words. Add captions in a normal editor afterwards. This is general practice, not a statement from Google.
• Do not impersonate. Do not use AI video of a real person, a real brand or a real event to make something look like it happened. Besides being unethical, it can cause legal trouble and is exactly how deepfake scams work.

Scam angle. Free AI video is a favourite lure for "video generator" apps that ask for a card number, or browser extensions that ask to read every page you visit. Use Vids from vids.new while signed in to your own Google account, and do not install third-party "AI video" extensions. Our article on malicious fake AI browser extensions shows how those traps work.

Turn a drawing into an app: the Gemini sketch trick

What happened. Also on November 5, 2026, Google published "Transform a quick drawing into a live app with Gemini." The method it describes is simple. You upload a photo of a hand-drawn sketch, such as an app mock-up, a habit tracker or a game board. You choose the Canvas tool and ask Gemini to code the app, with a prompt as plain as "Can you code this app?" Gemini reads the drawing and writes the code. You then test the result in a preview window and can ask for changes to colours, fonts or features.

What the page did not say. The announcement does not state which regions, devices or subscription levels can use it, and it does not mention limits on sketch complexity or usage. So if the option does not appear for you, that may simply be a rollout or plan difference. Do not pay for a plan just to get it before checking Google's live page.

Who this is for. A teacher who wants a small quiz game, a parent building a chore chart, a small business owner who wants a one-page price calculator, or anyone who learns by tinkering. It is a toy-sized builder, and it is good at that.

What to be careful about.

• Do not draw anything confidential. A sketch of your real client list, your login screens or your floor plan is data you are uploading. Treat it like any other upload to a free chat tool.
• Generated code is not audited code. An app that looks right in the preview can still contain mistakes. It is fine for personal use and brainstorming. If a "quick app" will handle customer information or payments, ask a professional to review it first.
• Do not paste secrets into the prompt. If the app needs an API key or a password, do not put it in the chat. Keep secrets out of AI conversations entirely.
• Check where the app lives. The announcement describes a preview. It does not say the app can be hosted or shared publicly, so do not promise anything to customers based on it.

If you are curious about how Google's other AI products fit together and which are free, our best free AI tools guide gives the wider map.

Claude Fable 5.2 and Mythos 5.2: what the new Anthropic models mean for normal users

What happened. On November 9, 2026, Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2. According to Anthropic's announcement page, they are its most capable models for demanding technical work. The page says Fable and Mythos are the same underlying model. Fable includes strong cybersecurity and biology safeguards, so all users can access it. Mythos is the version with fewer restrictions, and it is limited to vetted organizations through verification programs (with Claude Enterprise customers able to use it through a product called Claude Security).

What it costs and where it runs. Anthropic lists API pricing of $8 per million input tokens and $40 per million output tokens, with a faster mode at $16 and $80. It says Fable 5.2 is available on all platforms, including Amazon Web Services, Google Cloud and Microsoft Azure. Again, these figures are for people building products. A regular user pays a subscription, not per token.

What matters to subscribers. The announcement says eligible paid plans get extra Fable usage through a promotion, and that Fable is included for up to 50% of weekly usage limits on Pro plans and standard Team seats. Anthropic also says medium effort is the default in Claude apps and Claude Code, and it still recommends Claude Opus 5.5 as the everyday model, with Fable or Mythos for the most ambitious work. In plain terms: for emails, summaries and homework help, Anthropic itself says you do not need the biggest model.

Safety notes in plain language. Anthropic says Fable's classifiers block high-risk cyber activity (such as penetration testing and exploit generation) and most research biology, and that blocked requests are answered by a different, older model instead. It also says that outside enterprise zero-data-retention arrangements (a private beta for eligible customers), other customers have 30-day retention for safety monitoring. If you use Claude for personal or business work, that retention line is worth knowing: chats you send may be kept for a period for safety checks.

Claims we are not repeating. The announcement contains impressive benchmark numbers and research claims (for example about coding and science). These are the company's own, we cannot independently confirm them, and they do not change anything you do on Monday morning, so we leave them out.

What you should do. If you already use Claude, open the model picker and see whether Fable 5.2 appears on your plan. If you are new, our Claude desktop app beginner setup guide walks through installing it safely from the official download page. If you are choosing between assistants, our ChatGPT vs Claude vs Gemini comparison is the evergreen reference.

Scam angle. Anthropic's page mentions a Cyber Verification Program for Mythos. A scammer's version of that is a "request Mythos access" email asking for your company details or a payment. Real access for vetted organizations goes through the vendor's own site. A cold message offering "early access to the most powerful AI" is a phishing pattern. See our guide on how to spot AI-generated phishing emails.

Gemini Intelligence on Android and the November Pixel Drop

What happened. On November 10, 2026, Google's Android team published an update on Gemini Intelligence, its set of Gemini features built into Android devices, and Google separately announced the November Pixel Drop. The Android post lists several features:

• Rambler turns spoken thoughts into organized text in real time, through Gboard, on the Pixel 10 family and the motorola signature 27.
• Task automation handles things like ordering food, hailing rides and booking reservations across apps, and is expanding to the UK, India and Japan.
• Create My Widget builds a custom home-screen widget from a plain-language description, pulling information from Gmail, Calendar and Search. It rolls out over the next few weeks.
• Device Assistance lets you ask Gemini to free up storage, review which apps can access your location, or silence distractions while keeping important calls. Google says it needs Android 17 or later on select devices.
• Proactive suggestions on Android Auto are described as "coming soon" on Pixel phones with Gemini Intelligence.

Google says features are available in select countries and languages, that task automation and Create My Widget are for users 18 and over, that some features may require a subscription for higher usage limits, and that availability varies by device. The post lists the Pixel 10, 10 Pro, 10 Pro XL and 10 Pro Fold for the November drop, plus earlier devices including the Pixel 11 and recent Samsung models, and says Motorola's signature 27 is available for U.S. pre-order on November 19.

Why it matters for privacy. This is the item in the roundup that touches the most personal data. The post itself has no dedicated privacy section, and it notes that Create My Widget draws on Gmail, Calendar and Search data, and that Android Auto suggestions connect information across apps. That is how these features work: the more an assistant can see, the more helpful it can be, and the more you are trusting it with. Before you enable anything, ask yourself three questions. What does this feature read? Where is the result stored? Can I turn it off later?

What you should do. If you own a supported phone, treat each new feature as opt-in. Do not enable task automation for apps that hold your money (banking, trading, payment apps). Use the Device Assistance idea of reviewing which apps can see your location, even without the AI: Settings, Location, App permissions. If you do not have a supported device, there is nothing to do. This is a good example of "news" that does not apply to everybody, and that is fine.

Related reading. When an assistant can act across your apps, the risk is not only privacy but also being tricked into the wrong action. Our guide to AI browser agents and their security risks explains the pattern.

Adobe and Google: mostly announcements about the future

On November 10, 2026, Google published a post about Adobe's creative and productivity tools coming closer to Google's everyday apps, announced at Adobe's MAX event. The post lists a direct path from Google Drive and Gmail to Adobe Acrobat (an "Edit with" option in file previews), expanded Adobe connectors in Gemini and Gemini Enterprise, Adobe's creative tools coming to AI Mode in Search, and early exploration with Google DeepMind of a new image model for Photoshop. It says the Drive and Gmail pathway is coming "soon" and gives no other timelines. The DeepMind work is described as exploratory.

The post contains no privacy or data-handling details. For regular people the lesson is to wait. When these connections arrive, they will likely ask your permission to link two accounts. Read the permission screen. Granting an AI tool access to your Drive means it may read your documents, so link only when you know what the connector can see and can revoke it later in your Google account's security settings.

OpenAI this week: businesses, veterans and caregivers

OpenAI's news feed lists several posts this week. Three of its pages returned an access error to our tools, so what follows is limited to the short descriptions in OpenAI's own feed, and we encourage you to read the full posts yourself.

• "Introducing Businesses in ChatGPT" (November 10). The feed describes it as a way to "manage your business information, get insights about how your business shows up, and turn interest into action in ChatGPT." We could not read the full post, so we cannot tell you the price, the countries or the setup steps. If you run a local business, read the post before you let any tool edit how your business is described, and keep your own copy of your opening hours, address and services.
• "ChatGPT Plus for veterans and servicemembers" (November 11). The feed says over 20,000 former and transitioning military personnel have redeemed a no-cost year of the paid ChatGPT tier. We could not confirm eligibility rules or countries from the full page. Do not click offers that claim to give "free ChatGPT Plus" to anyone: scam pages copy real promotions. Go to OpenAI's own site.
• "Helping caregivers use AI in everyday life" (November 10). The feed says OpenAI Academy is running free ChatGPT training sessions in five U.S. cities. It is a U.S. program, so it is not something to act on from Quebec, but it shows the trend: companies are now running AI literacy programs for everyday users.

The feed also lists enterprise items (ChatGPT Enterprise in the AWS Marketplace and several customer stories). They are aimed at large organizations and do not change anything for a household or a small shop.

Microsoft: nothing new for regular users in the window

We checked Microsoft's official blog feed. Its most recent items were dated October 28, 2026 or earlier, and the ones about Copilot were about developer and enterprise tooling, such as a Copilot harness in Microsoft Foundry. We found no consumer-facing Copilot announcement dated November 4 to 11, so we are not inventing one. If you use Copilot at work, our guides on Microsoft Copilot for business and Copilot Cowork explained for small business are still current.

Slightly older Gemini app changes you may have missed

The Gemini app release notes we opened list no entry dated in November. The latest one we read is dated October 13, 2026: Gemini 4 Argon rolling out to Google AI Pro and Ultra subscribers, Antigravity built into the Gemini app for macOS, and a "Personalize" menu that lets you tailor how Gemini responds, add skills and connect more than 40 apps. An earlier entry dated September 30 says Skills will soon replace Gems and that Gems will be migrated automatically, with no date given. Our tool could not read the entire page, so there may be more entries. The takeaway is the same: if you rely on Gems, expect a change and check the release notes occasionally. Google's Flow help page says Veo 3.1 Fast, Veo 3.1 Quality and Omni Flash 1.1 retire in Flow on December 5, 2026, and the Veo 3.1 preview models in the Gemini API shut down on October 22, 2026. Details in our Veo guide.

The scam and privacy side of this week's AI news

Every big launch creates a wave of opportunists. Based on how scam campaigns usually work (this is a general pattern, not a claim about a specific incident this week), you can expect the following in the weeks after announcements like the ones above.

Voice cloning deserves special mention, because better video and audio models make it easier to fake a voice or a face. Agree on a code word with your family and your accounting contact. We cover business-side defences in AI voice cloning scams and deepfake audio. If you want a guided session, our AI scam shield session teaches you to spot these in under an hour.

Privacy basics that apply to every item above

• Free tools are not private vaults. Vendors can keep chats for safety monitoring, and some use free-tier content to improve products. Anthropic's announcement mentions 30-day retention for safety monitoring for most customers; Google's own help pages describe human review of a subset of chats. Read the setting in your account rather than assuming.
• Connectors widen the blast radius. The more apps you link (Drive, Gmail, Calendar), the more an assistant, or a trick played on it, can reach. Link the minimum.
• Sensitive categories stay out. Health, legal, banking, immigration and children's information do not belong in a general consumer chat.
• Review what is saved. Once a month, delete old chats you no longer need.
• Use strong sign-in. Turn on two-step verification for every AI account, ideally with a passkey or authenticator app.

Our full AI privacy and safety checklist goes through the settings step by step, and if you run a team, the problem of employees using tools nobody approved is covered in shadow AI: the hidden risk.

Your weekly AI news checklist

Three realistic scenarios (illustrative)

Realistic scenario 1: the curious parent. Nadia, a parent in Laval, hears that Gemini can turn a sketch into an app. She draws a chore chart on paper, uploads a photo, uses Canvas and gets a working preview in a few minutes. She does not include her children's names or school details. That is exactly the right level of risk: a harmless drawing, a toy result.

Realistic scenario 2: the small shop owner. Marc owns a four-person plumbing shop. He sees "Businesses in ChatGPT" in the news. He reads OpenAI's post before touching anything, writes down his current hours and service area in a document, and decides to wait a few weeks until the details are clear. Waiting cost him nothing; a rushed edit might have changed how customers see his business.

Realistic scenario 3: the near miss. Julie, an office manager, receives an email saying her team has been selected for "early access to the newest AI model" and asks her to sign in with her work account on a linked page. She notices the sender address does not match the vendor, does not click, and forwards it to her IT provider. The pattern is classic credential phishing dressed in AI news.

What this week's AI news costs you (CAD view)

For regular people the honest answer is zero. The Google Vids video allowance is free with a Google account (three 15-second videos a month), the sketch-to-app feature was described without any pricing, and Fable 5.2 appears on Claude plans you may already have. The developer prices we quoted ($2 to $4 per million input tokens for Gemini 4 Carbon, $8 for Claude Fable 5.2) are in U.S. dollars as shown on the vendors' pages and only matter if you pay per use. For comparison, a one-hour setup session with a technician is a single fixed fee, and our expert consultation is 119.99$ CAD for 60 minutes. If you want to avoid trial and error, that is often cheaper than a month of a plan you do not need.

For small businesses: turning news into a plan

If you run a business, the week's news suggests three habits. First, assign one person to watch official vendor pages once a week rather than letting everyone chase headlines. Second, write a one-page rule about what staff may paste into AI tools, and keep a list of approved tools (our guide on writing an AI usage policy has a template approach). Third, look for one repetitive task, such as quoting, scheduling or follow-up messages, that a tool can handle safely. If you want help with the marketing side, for example AI automation for Quebec contractors and local businesses from Lead4Pro, a Quebec marketing agency, is a place to start. For the security and setup side, IT Cares can configure accounts, privacy settings and two-step verification for you, and our AI services shop lists fixed-price sessions.

When to call a pro

Call a technician if you clicked a link from an "AI launch" message and entered a password, if an AI extension asked for broad permissions and you accepted, if a family member received a suspicious call in a familiar voice, or if you want a new assistant set up with sensible privacy defaults on all your devices. IT Cares offers remote support across Canada and on-site visits in Quebec. We will not sell you a tool you do not need, and we are not affiliated with any of the vendors named here. You can book a session or call 1 (888) 711-9428 and we will connect to your device while you watch.

Official resources to bookmark

• Anthropic news: anthropic.com/news
• Google blog (Gemini and AI): blog.google
• Gemini app release notes: gemini.google/release-notes
• OpenAI news: openai.com/news
• Canadian Anti-Fraud Centre: antifraudcentre-centreantifraude.ca
• Office of the Privacy Commissioner of Canada: priv.gc.ca

Frequently asked questions

Related guides

• Free Google AI tools list 2026
• Claude desktop app beginner setup guide
• Google Disco browser: what it is and how to use it
• Google AI video for free
• AI privacy and safety checklist
• Best free AI tools 2026
• IT Cares AI services shop

Sources and official references

Last verified: November 11, 2026

• Anthropic news page (opened 2026-11-11)
• Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2, Nov 9, 2026
• Google: Gemini 4 Carbon is now available, Nov 10, 2026
• Google: Omni 1.2 Flash in Google Vids, Nov 5, 2026
• Google: Transform a quick drawing into a live app with Gemini, Nov 5, 2026
• Google: Gemini Intelligence, new features on more devices, Nov 10, 2026
• Google: Adobe and Google partnership integration, Nov 10, 2026
• Google blog feed (opened 2026-11-11)
• Gemini app release notes (opened 2026-11-11)
• OpenAI news feed (opened 2026-11-11)
• Microsoft official blog feed (opened 2026-11-11)