Microsoft's January Terms Update: A Rebrand, a Bundle, and a Licensing Change That Matters

Microsoft opened the year with a Product Terms update that reads like housekeeping and is not. Three items are worth the attention of anyone running workloads on Azure or embedding models in a product.

Another rename

Azure AI Foundry is now Microsoft Foundry, and what were Azure AI Foundry Models are now Microsoft Foundry Models. That is the second rebrand of the same platform inside twelve months. The platform underneath is unchanged; the cost lands on everyone maintaining procurement documents, internal runbooks, and onboarding material — that work now gets done twice.

Practical advice: reference these services by function in your internal documentation, not by marketing name, and keep the vendor name in one place you can update once.

The substantive part: model licensing

The same update introduced licensing terms covering third-party AI models offered directly through Azure. If your product calls models through Azure, this is the section to actually read rather than skim. What a model may be used for, how outputs may be handled, and which terms flow through to your own customers are contractual questions, and they differ by model provider even when the API surface looks identical.

We have seen teams assume that because a model is available in a vendor’s catalogue, the vendor’s standard terms govern it. Frequently they do not.

Security Copilot arrives whether you asked or not

From early January, eligible Microsoft 365 E5 tenants had Security Copilot capacity provisioned automatically — 400 Security Compute Units per 1,000 users, scaled to tenant size, with no Azure setup required. Zero-click provisioning is convenient, and it also means capability appeared in tenants whose security teams had not planned for it.

• Audit what is now enabled rather than discovering it during a review.
• Understand the SCU model — units reset monthly and are shared across workspaces, so consumption patterns matter.
• Decide deliberately whether to build process around it or leave it dormant. Both are valid; drifting is not.

The backdrop

Microsoft reported Azure and other cloud services growth of 39% year over year in its FY26 Q2 results in late January. That growth is what funds the AI infrastructure spend, and it is also why bundling moves like the Copilot inclusion keep arriving — packaging AI capability into existing suites is the fastest route to monetizing it.

Expect more capability to arrive pre-provisioned rather than purchased. That shifts the work from procurement to governance.

What we tell clients

Read the terms updates for the licensing sections and ignore the renames. Keep your architecture’s model layer replaceable so that a provider’s terms change is a configuration decision rather than a rewrite. And review what your tenant has quietly gained each quarter — the surface area of an enterprise agreement now moves without you doing anything.