September 30, 2026

Latest

Microsoft 365 Family Copilot Leaves Five Out

Microsoft sells its Family plan as one subscription for up to six people, and since early 2025 it has sold that plan with Copilot built into Word, Excel and PowerPoint. Now open Word on your partner's laptop. They're signed in with their own account, on your plan, and the Copilot button asks them to buy something. That is Microsoft 365 Family Copilot as it actually behaves: the price covers the household, the AI covers whoever pays the bill.

TL;DR: On a Family plan, Copilot inside the Office apps works for the subscription owner only, and no consumer upgrade changes that.

  • The five people you share with keep the apps and their own storage, but not Copilot in Word, Excel, PowerPoint or Outlook.
  • Premium raises the owner's Copilot allowance and adds no AI for anyone else.
  • The only way to give a member Copilot in the apps is their own Personal plan.
  • An owner who barely uses Copilot should try cancelling first, because Microsoft may offer a cheaper Classic plan without AI.

Why does Microsoft 365 Family Copilot stop at the owner?

Microsoft licenses the AI features in Word, Excel, PowerPoint, Outlook and Designer to the person who pays, so family members share the apps and storage while the in-app AI answers to the owner's account alone.

Microsoft's own Copilot FAQ for Family and Personal subscriptions says it without hedging: those AI features "are only available to the subscription owner and cannot be shared with others" (retrieved 30 September 2026). At the kitchen table that reads simply. Your teenager drafting an essay and your partner building a budget both see the button. Neither can use it on your plan.

I think this is the most wasted opportunity in Microsoft's consumer lineup. Family puts Office in front of people who would never pay for it, exactly the audience Copilot needs to win. Members get the free Copilot chat app instead, and nothing inside their documents. The throttling logic is familiar from what a $20 AI chatbot plan buys before it throttles; here the limit isn't how much you use, it's who you are.

Designer is on the same list, so a member making a school poster should look at our Figma AI, Adobe Firefly and Canva Magic Studio comparison instead. Why Microsoft drew the line at the owner isn't documented. My read, and it is only opinion, is a compute-cost decision dressed as a licensing rule.

Four numbers decide whether the rule is an annoyance or a bad deal. NBC News reported in January 2025 that Microsoft raised consumer prices as Copilot entered the apps; Office Watch reported the Family plan's monthly AI credits in March 2026. The last cell is my own division of the increase by a year of credits.

Years Without a Price Rise

12 years

Ended when Copilot arrived

Family Plan Increase

$30 a year

Billed for AI one person uses

Owner's Monthly AI Credits

60 credits

The whole household's allowance

Price per Credit

About 4 cents

Only if the owner spends all

The credit allowance bites hardest. It isn't a pool split six ways. It belongs to the owner and resets monthly, so a family picturing Copilot helping with homework is really looking at one person's quota.

"

Every sharing household paid for Copilot's arrival. Only the person on the bill can open it.

Family or Premium: which plan fits a shared household?

Family suits most shared households, because Premium only raises the owner's own Copilot allowance; neither plan lets the other five members use Copilot inside Word, Excel, PowerPoint or Outlook.

The usual forum advice is to step up to Premium. It isn't a fix: Microsoft's pricing page prints the same owner-only footnote under both plans.

Dimension Family vs Premium What it means for you
💰 Yearly price Family $129.99 a year
Premium $199.99 a year
⚠️ $70 more buys the owner more use
💰 Cost per seat Family $21.67 each, six seats
Premium $33.33 each, six seats
✅ Office stays cheap when split
🔋 Storage Family 1 TB per person
Premium 1 TB per person
✅ Members lose nothing on files
📊 Owner's AI use Family fixed monthly credits
Premium extensive, no monthly cap
⚠️ Worth paying only past the cap
🔒 AI for members Family none in the apps
Premium none in the apps
❌ Upgrading adds no Copilot seat
🛠 Member fallback Family free Copilot chat
Premium free Copilot chat
⚠️ Help sits outside the document
🧾 Cheaper exit Family Classic, $99.99 if offered
Premium drop back to Family
⚠️ Classic tends to surface at cancel
🏁 Best suited for Family owner uses it weekly
Premium owner drafts daily
🏁 Neither, if a member needs Copilot

Premium earns its extra cost in one case: the owner runs out of credits most months. For anyone else, the route to Copilot in the apps is Microsoft 365 Personal, listed at $99.99 a year for one person.

Owner. Copilot on. Member. No Copilot. Member. No Copilot. Member. No Copilot. Member. No Copilot. Member. No Copilot.

Copilot in the apps for all six people costs $599.94 a year through six Personal plans, about 4.6 times the Family price, so buy it only for the member who would really use it. Derived from Microsoft's US list prices, retrieved 30 September 2026.

What catches sharing households out

The owner-only rule catches households twice: at renewal, when an AI-era price arrives for a plan where one person gets the AI, and at homework time, when a member finds the Copilot button asking them to subscribe.

No consumer tier fixes it. Even Microsoft 365 Pro, at $99.99 a month and pitched on "highest usage", carries the same owner-only footnote. At that price, a plan sold to six people should at least let the owner assign seats.

The cheaper exit isn't advertised. Office Watch reported in January 2026 that Classic plans, which drop Copilot, "might be offered to people who try to cancel". It's a retention offer, not a menu item. Members keep the free Copilot app, and the AI browsers we compared against Chrome cover a lot of that ground for nothing.

  • A member's Copilot prompt in Word leads to a purchase page, not to your plan.
  • Classic removes Copilot for the owner too, so check your own use before asking for it.
  • A member who buys Personal while staying on your plan pays twice for apps and storage, so free up their seat.

Check these before your renewal date

  • You opened Copilot inside Word or Excel last month. If not, ask about Classic.
  • You hit the credit limit more than once this year. Only then does Premium pay.
  • A member works in Word or Excel most weeks and has asked about Copilot. Move them to Personal.

This week, before the plan renews, ask the household one question: who actually wants Copilot inside their documents? If it's one person, move them to their own Personal plan. If it's only you and you rarely touch it, start a cancellation and see whether Classic appears. Either way, stop paying for Premium on the theory that it shares.

September 21, 2026

AI Chatbot Usage Limits: What Your $20 Buys

The plan card promised expanded messages. What it delivered, four hours into a Tuesday, was a grey banner and a wait. Nothing on the page you paid from said how many messages you had, or when they came back. That is the shape of AI chatbot usage limits in 2026: the price is a number, the allowance is an adjective.

Subscription plan cards showing AI chatbot usage limits as multipliers instead of countable numbers

Key Takeaways: only one of the three big vendors prints a number you could check a bill against.

  • GitHub counts Copilot in premium requests and publishes the monthly number.
  • Google sells multiples of a base it never publishes, so a higher tier stays unknown.
  • OpenAI describes its allowance in adjectives and hides the real limits behind a link.
  • Size any upgrade against your heaviest week; the average week never causes the lockout.

Why do AI chatbot usage limits stay unpublished?

Because an unpublished ceiling can move: vendors size capacity week to week, and a printed number becomes a promise they must honour or explain, while an adjective can be quietly re-tuned overnight.

Look at what OpenAI's pricing page commits to. Plus is sold as expanded messages and uploads, with "Limits apply" set as a link rather than a figure. Free is "unlimited text chats, subject to abuse guardrails". So, in almost identical words, is Pro. When the cheapest tier and the dearest carry the same qualifier, the qualifier has stopped meaning anything.

This is not new, and not confined to chat. It showed up in the AI browsers that promised to replace Chrome, and it runs through the Figma AI versus Adobe Firefly comparison as well: a monthly price, a vague noun, and the ceiling written where no buyer looks.

Google is the interesting middle case, or the frustrating one, depending on whether you are comparing or buying. Its subscription page publishes precise prices and deliberately imprecise allowances: every paid tier is a multiple of a free tier whose own number is never stated. The four figures below are the ones that decide whether a plan fits your month, and they come from Google's subscription page and GitHub's billing documentation, both read on 21 September 2026.

How Often Gemini Refills

Every 5 hours

Weekly ceiling still ends the week

Copilot Cost Past The Cap

$0.04 per request

The only unit price anyone prints

Copilot Pro+ Monthly Allowance

1,500 requests

A ceiling you can actually count

Gemini's Top Multiplier

20x

Best rate, worst entry price

The refresh window is the figure people misread. A rolling reset sounds forgiving, because something comes back several times a day, but it sits under a weekly ceiling that does not roll. Spend hard on Monday and the window still opens on Thursday with nothing left to refill it. It is the design, not a bug, and it rewards flat workloads.

"

Four cents a request is the only unit price any of these three vendors will print. Everything else you buy is a multiple of a number they have chosen not to publish.

What twenty dollars a month actually buys

At the twenty dollar mark you are buying either a multiple of an unpublished number or nothing published at all, because the only vendor here that prints a countable allowance does not sell a tier at that price.

Two ladders, side by side. Every figure below comes from Google's subscription page or GitHub's billing documentation, except the break-even in row six, which is ours. OpenAI has no column because its plan cards carry no allowance figure to put in a cell.

DimensionGoogle Gemini plansGitHub Copilot plansWhat it means for you
Entry paid tier$4.99 a month$10 a monthCheaper to start, not cheaper per unit.
What it buys2x the free tier, base unpublished300 premium requests a monthOne number divides into your week; one cannot.
The $20 tier$19.99 for 4x free usageNo tier at this priceFour times the entry price, twice the multiplier.
Top consumer tier$99.99 for 5x the $20 tier$39 a month, the individual ceilingOnly one vendor sells headroom above $40.
Running overYou wait for a refreshBilled per extra request, work continuesWaiting costs hours, billing costs forecastable money.
Upgrade break-evenNot computable, base unpublishedAbout 1,025 requests a month, oursYou can time an upgrade on one side.
Best Suited ForEven weeks, and patience during a lockoutAnyone who must explain the AI billOnly one column gives a number before you pay.

That break-even is ours, not GitHub's, and it is the only upgrade decision here you can calculate rather than guess. Take the gap between the two Copilot tiers, divide by the published overage rate, and you land near 1,025 requests a month. Under that line, pay the overflow. Over it, upgrading pays for itself.

20%. 40%. 40%. Countable number. 2 of 10 tiers. Relative multiplier. 4 of 10 tiers. No figure at all. 4 of 10 tiers.

Across the ten consumer tiers these vendors sell, only two state an allowance you could check a bill against. Counted by hand from the pricing pages named above, read on 21 September 2026.

Do AI plan limits reset daily or weekly?

Both, and that is the problem: Google refreshes on a rolling window that sits inside a weekly cap, GitHub resets on your monthly billing date, and OpenAI publishes no reset schedule you could plan a working week around.

There is a second catch buried in the Copilot figures. GitHub's billing documentation scopes those published request counts to subscribers on an existing annual plan who stayed on legacy premium-request billing after 1 June 2026. The count is real and checkable. It is also grandfathered, so a buyer signing up today may not be on that plan.

My own read, and this is a stance rather than a finding: the vague multiplier is less a transparency failure than a capacity hedge. A vendor that cannot predict its own inference costs will not print a ceiling it may have to defend. Understandable. It still leaves the buyer holding the uncertainty, the same trade as when courts cleared AI shopping agents to buy on your behalf: the capability shipped, the accountability did not.

  • A weekly ceiling under a rolling reset means a heavy Monday can cost you Thursday.
  • A multiple of a base that quietly shrinks is a cut nobody has to announce.
  • Check which plan your signup date put you on, not the one the documentation describes.

Four things to check about your own month before upgrading.

  • You can name your plan's allowance without opening a help article.
  • Your heaviest week runs roughly double your average, not barely above it.
  • A lockout at three in the afternoon would cost you something real.
  • Somebody other than you sees this invoice and will ask about it.

So do this before your next billing date. Open the limits page your plan links to, write down the one number or multiplier it gives, and divide it by your heaviest week. If it will not divide, because there is no number, you have learned the most useful thing about that plan: you are renting an adjective, not buying an allowance. Decide on that, not on the plan card.

Related: why Microsoft 365 Family Copilot works only for the plan owner

September 4, 2026

How to Use DigiYatra Now That It's Mandatory

You land in Delhi before dawn, connecting onward to Singapore, and the officer hands your passport back without reaching for a stamp. The camera above the desk already knows your face. Nobody sat you down and explained how to use DigiYatra, and at that terminal it has stopped being a choice you get to make. The queue moves faster. Your record of having been there moves somewhere you cannot see.

How to Use DigiYatra Now That It's Mandatory
DigiYatra's face scan became compulsory for international passengers at Delhi, Mumbai, Bengaluru and Hyderabad on 1 June 2026. Enrolment needs an Aadhaar-verified selfie plus your boarding pass inside the app. Domestic travel stays opt-in. Foreign passport holders without Aadhaar have no published enrolment route.

Why DigiYatra, and why the rush?

India's Bureau of Immigration stopped stamping international boarding passes on 1 September 2026, so the paper trail a traveller once carried home now lives entirely inside an app and an airline database. That is the real story, and it broke a week ago with almost no coverage joining the two changes together.

Two things happened this year, three months apart, and they only make sense read as one move. First the face scan became the default way through the terminal at the busiest international gateways. Then the ink stamp, the one artefact that proved you physically left and re-entered the country, was withdrawn. Visa officers asked for that stamp. So did tax officers, and employers running background checks. Now the answer is a PDF and a database query.

What is DigiYatra actually doing at the gate?

It matches a live camera image against an encrypted template built from your Aadhaar-verified selfie, then releases your boarding record to the checkpoint. Outlook Traveller reported the operational detail plainly: the selfie and the boarding pass go into the app before you reach the airport, and the gate does the rest. There is no counter and no second look. This is the same pattern as India's Android developer verification rules, where a convenience is announced first and the compulsion arrives later, quietly, in a circular.

The programme's governance is the part worth sitting with. DigiYatra Foundation is a private not-for-profit, and because it is private it sits outside the Right to Information Act, 2005. So the body holding a biometric template of every international flyer through four major airports cannot be asked, by an ordinary citizen, how long it keeps that template. The Internet Freedom Foundation has raised exactly this: privacy, surveillance, exclusion errors, and no institutional accountability. A Kerala High Court PIL alleging commercial misuse of passenger data is still live. None of that stopped the mandate.

Upload lead time

48 hours

Before scheduled departure

Passenger fee

Rs 0

Free to enrol and use

Survey responses

21,000

LocalCircles panel, January 2024

Enrolled unknowingly

29%

Of signed-up flyers, as of 2024

That last figure is the one that should shape how you treat the app. It was measured while enrolment was still sold as voluntary, and it describes people who handed over a document at a counter and were signed into a biometric programme without registering that it had happened. A system with that much accidental enrolment in its optional phase does not suddenly grow better consent hygiene once it becomes compulsory. It just stops needing consent.

"

Nearly a third of enrolled flyers never chose the face scan. That was the voluntary era. It is now the rule at four gateways, and consent has stopped being part of the conversation.

How to Use DigiYatra: What Each Step Demands

You enrol once in the DigiYatra app with an Aadhaar-verified selfie, attach each boarding pass to that profile before you travel, then walk to the e-gate where a live camera matches you against the stored template and opens the checkpoint.

The steps are simple. What sits behind them is not, and the table below is the version nobody prints on the airport signage. Read the Identity row first if you hold a foreign passport, and the Paper trail row before your next visa application.

Category Detail Insight
Scope Transit passengers included, not only those departing India Connections through India now need enrolment
Identity Aadhaar-verified selfie only, no published passport-only route Foreign passport holders hit a wall
Ownership DigiYatra Foundation: AAI holds 26 percent, five airport operators 14.8 percent each Private body, no RTI questions allowed
Paper trail Ink stamp withdrawn, e-boarding pass is now the only travel record Save the PDF before you fly
Gate fallback Budget 15 extra minutes when a face match fails at the e-gate Manual desks stay open, queues return
Battery floor Reach immigration above 30 percent charge, or carry one printed pass A dead phone is a missed gate

The Ownership row is the one that changes how you should read every reassurance about deletion. A shareholding split between the state airports authority and the operators it regulates is a commercial arrangement wearing safeguard language, and the entity it created answers to neither a regulator nor a citizen with an RTI form. The 15-minute buffer in the Gate fallback row is not published anywhere either. It comes from the plain logic of the system: if the automated lane rejects you, you rejoin a manual queue that was sized for fewer people than it used to hold.

1 2 3 4 Install app · Aadhaar selfie · Attach boarding pass · Walk the e-gate One time · One time · Every trip · Every trip

The enrolment flow in four steps, as the DigiYatra app itself sequences it: the first two are done once, the last two repeat on every international trip.

Can you register for DigiYatra without a boarding pass?

You can create the profile without one, since enrolment needs only the Aadhaar-verified selfie, but the boarding pass has to be added to that profile for every single trip before the gate hardware will recognise you and let you through.

Which means the app is two systems wearing one name. The identity half is permanent and lives with a private foundation. The travel half is per-journey and expires. Most of the friction people report sits in the second half: a pass added too late, an airline that issues the document only at the counter, a name spelt differently on the ticket and the Aadhaar record. The rollout is fine, mostly, if your face matches on the first attempt.

The camera is the least interesting part of this. What nobody has published is a retention window for the biometric template, and the body holding it is structurally unanswerable to the people it holds it on. That is a stance, not a finding, and it is the same gap that shows up whenever the rules land before the plumbing does, exactly as they did with AI shopping agents and India's missing payment rail.

  • Download and store the e-boarding pass as a PDF the moment it arrives, because the immigration stamp that used to prove your trip no longer exists.
  • Check that the name on your ticket matches your Aadhaar record character for character, including initials, before you add the pass.
  • Ask for the manual counter if you do not want the scan. It still exists at these terminals for people the automated lane rejects, and staff will point you to it if you insist.
  • Keep a printed pass in your bag. It looks absurd next to a face scanner, and it is still the cheapest insurance you can carry.

Key takeaways before your next international departure

  • Enrolment is one-time, the boarding pass step is not. Treat it as part of check-in, not part of setup.
  • Your proof of having travelled is now a file you control. Nobody at the airport is keeping a copy for you.
  • Non-Aadhaar travellers have no documented path. Plan for the manual lane and the time it costs.

Do one thing this week. Open your last international e-boarding pass, save it somewhere that survives a lost phone, and do the same for every trip from here. The face scan is settled and arguing with it at the gate will only cost you your connection. The paperwork is what you can still control, and the day a consulate or a tax officer asks you to prove you left the country, that saved file is the whole of your answer.

Keep reading

Home