For years the deal looked simple: OpenAI built the models, Microsoft sold them. That era is closing. Microsoft now ships its own Microsoft AI models, and its Copilot products increasingly let you pick which engine answers you. If your business runs on Microsoft 365, that change reaches your desk sooner than any headline about superintelligence.
What Microsoft AI models actually are now
Mustafa Suleyman formed the MAI Superintelligence Team in late 2025. In April 2026 the group shipped three in-house models for speech, voice and images. Then, at Build in June, it unveiled a family of seven models built from scratch. The flagship is MAI-Thinking-1, a reasoning model Microsoft claims draws level with Anthropic’s Claude Sonnet 4.6 in blind human testing. Suleyman framed the goal plainly: “long term self-sufficiency for Microsoft and our partners.” GeekWire’s report from Build walks through the full line-up.
Copilot is becoming multi-model
Here is the part that touches your subscription. Microsoft 365 Copilot no longer runs on one supplier. Anthropic models sit alongside OpenAI ones, and users can choose per task. Meanwhile Microsoft ended its 2026 financial year with roughly 30 million Copilot seats, so this is not a pilot. In practice you gain choice and inherit a question: which model saw your document, and under whose terms? We looked at the value side of the product in our review of Copilot for small businesses.
Three consequences for a smaller company
- Lock-in loosens, slightly. A vendor that builds its own models can drop a partner’s, so avoid workflows tied to one exact model name.
- Your data map changes. New engines mean new subprocessors, and your privacy notice should name the category, not the brand of the month.
- Quality now varies by task. Multi-model means testing, because the best drafting model rarely wins at spreadsheets.
Those are governance points, not technical ones. Consequently the fix is a document rather than a deployment. Our piece on what AI safety actually means for your business covers the questions worth settling in writing.
What to do this month
Start with an inventory. List every AI feature your team already uses, then note which vendor and which model sits behind it. Next, pick the two tasks where AI saves real hours and test the available models against each other on your own work. Finally, write down the switching cost for each tool. That number tells you whether choice is a benefit or a trap. Microsoft AI models will keep multiplying, so make that review a habit rather than a one-off. If you are still deciding on a default, our guide to choosing an AI model and the wider Tech Industry archive lay out the trade-offs.
