Microsoft Starts Capping Free Copilot Chat for Companies With 2,000+ Seats, and Won't Say Where the Cap Is
Fair-use limits on more Copilot Chat features reach large Microsoft 365 tenants today, a week after smaller ones. Admins can't see or change the limits, and the only published number is the $30 licence that removes them.
TL;DR
Microsoft is applying fair-use limits to more Copilot Chat features for users without a Microsoft 365 Copilot licence. The rollout began on 2 November for tenants with up to 2,000 seats and starts on 9 November for larger ones. Microsoft does not publish the limits, says they vary through the day, and gives admins no way to view or change them. Users who hit a cap see an in-product message with a retry time, usually next to a button to request the paid licence, which costs $30 per user per month.
Copilot Chat is included in business Microsoft 365 subscriptions at no extra cost. As of today it comes with more strings attached. Microsoft's Message Center post MC1536163, published on 3 November, says it is applying "fair use" limits to additional Copilot Chat features and capabilities for people who don't have a Microsoft 365 Copilot licence.
What Microsoft has said
| Question | Answer from the Message Center post |
|---|---|
| Who is affected? | Users without a Microsoft 365 Copilot licence in commercial tenants |
| Where? | Copilot Chat in the web, desktop and mobile apps |
| When? | Tenants with up to 2,000 seats from 2 November; larger tenants from 9 November |
| What is limited? | Image generation was already limited. Microsoft says it is "applying this approach to additional" features and doesn't list them |
| How high are the limits? | Not published. They can change and can vary during the day |
| Can admins see or change them? | No. Microsoft sets and manages them |
| What does a user see? | An in-product notice saying when to try again. A response already being written is not cut off |
What the notices leave out
- The numbers. There is no message count, token budget or feature list, so a team can't predict whether a normal working day will hit a cap.
- The feature list. The post names image generation as an existing limit and says the approach now covers "additional" features, without naming them. It adds that limits may expand to more capabilities over time.
- A way to measure impact. Admins are told to brief the help desk, not given a report of who hit a limit.
The upgrade path
Itdaily reports that in most organisations the limit message includes an upgrade button. It sends a request for a Microsoft 365 Copilot licence to the admin, who approves or rejects it in the admin centre. Where self-service purchase is switched on, the user can buy a licence on the spot, and admins can disable that per product. Itdaily also reports that Microsoft tells partners a user hitting the limit is showing interest in the paid product, and that resellers can see in the partner portal which customers have users running into limits. The paid licence is $30 per user per month, with Microsoft 365 Copilot Business for smaller companies.
How we got here
- Since 15 April, Copilot inside Word, Excel, PowerPoint and OneNote has worked only with a Microsoft 365 Copilot licence, according to Itdaily. Unlicensed users have seen the label "Copilot Chat (Basic)".
- Licensed users were then given priority when capacity is tight.
- This week's change extends the same approach to more chat features for everyone without a licence.
What this means if you choose an AI assistant for your team
- Treat "included with Microsoft 365" as a trial, not a budget line. With no published limits, you can't forecast what unlicensed staff can do on a heavy day.
- Test it before you rely on it. Have a few people use Copilot Chat for a normal week and note when the retry messages appear, since the admin centre won't tell you.
- Compare on published limits. Tools that state message or usage allowances in their pricing let you do the arithmetic. That's a fair question to put to any assistant you're weighing against Copilot.
Sources
AI Industry Reporter
Priya covers model releases, industry announcements, and the gap between what labs claim and what independent evaluators actually find. She reads the primary source - the paper, the system card, the benchmark org's own statement - before writing a word.
More on AI Chatbots & Assistants
Thinking Machines' Inkling 2 Small: Scores 31 on Artificial Analysis, Ships Closed, Promises Apache 2.0 Weights Next Week
Priya Nair · 5 min
Anthropic's Amended IPO Filing: $16.96 Billion in Q3 Revenue, a $567 Million Operating Loss and $5 Billion of Short-Term Compute
Priya Nair · 5 min
Ai2's StateBench: Frontier Models Fall to 13% or Lower When Long Documents Keep Changing
Priya Nair · 4 min