MS offers a full competitive product, and they allowed that product to use features and capabilities that wouldn't be available to others. At their scale, you can't do that without drawing scrutiny from regulators. They could have at least kicked their own product out to userland, or just stopped offering one and focused on kernel stability the way Apple has.
Since this is all spitballing, perhaps they even suggested that the EU, and the EU said "no, it must be kernel access" and capitulated because this served their own interests while also letting them blame the EU for the lack of kernel stability.
The difference between XProtect and Defender, is that XProtect utilizes the same kernel interface that Apple provides to other EDR solutions, whereas Microsoft was / is only willing to implement Defender using an in-kernel solution.
As a result, it would be anti-competitive for Microsoft to deny that level of access to third party EDR solutions.
I do wish that the EU would offer a "grace period" for "platform owners" or "gatekeepers" that would allow them to use non-public APIs for their first-party solutions while they stabilize and work out the public API offering. I think there should be some requirements for them to "earn" that grace period, such as allowing interested third-parties to get early access and provide feedback on those APIs. Regardless, it takes time to stabilize those APIs, and the current EU approach of requiring all new APIs to be fully public on day one just isn't workable.
It also isn’t a separate product, it is bundled with the OS.
I genuinely want to understand this - how everyone else got it right and this entity got it wrong.
Well, both dominant mobile operating systems already enforce a very strict security model. Too strict if you ask me because on neither you as the person actually owning the damn thing can become root without anything in the system throwing wrenches in your way, but hey.
The result is, most people don't even need a virus scanner, and the blast radius of a bug in an application is usually limited to whatever data that application can access. It's been ages since I heard about the last major mobile OS malware in the wild that didn't involve one of the world's secret services or their civilian suppliers - and that way, one can argue, all the security theatre actually works: it's driven the price of successful attacks so high that only nation states can afford it, and even these have to restrain themselves to only use these exploits against targets of extremely high value.
Is it possible to rootkit phones? Yes; does a typical user do it - heck no.
iOS (and to a lesser degree, macOS) use notarization to be able to pull apps found of being risky as needed.
Additionally they implement much stricter levels of sandboxing so that even if something slips by all detection, it’s heavily limited in what it can do without a users explicit permission or a catastrophic vulnerability.
(There's a big android user population without access to the Android app store at all)