How so? Even if they don’t catch malware during notarization, Apple also reacts pretty quickly to invalidate a developer’s code-signing certificate if they use it to sign apps that contain malware (as soon as Apple is made aware of that malware-app, for which they maintain relationships with both major antivirus vendors and independent security researchers.) Your computer then receives the new Apple code-signing CRL in a silent update, and won’t run the app (or any app by that developer) any more. Even if you’re offline at the moment, and so can’t contact the notarization servers to find out the app has been denotarized, as long as you’ve been online at any point since the CRL was updated, you’ll be protected. (And where does malware come from? These days, 99% of the time, the network. So if you stay offline, you’re extremely unlikely to run into novel malware anyway. And if you’re online to receive the malware, you’re almost certainly going to have received the CRL update first.)

And sure, there’s a small period of vulnerability before Apple is made aware of new malware; but most malware infections are not from zero-day malware, but rather from malware that’s been going around for a long time already. (And I believe they also push ‘disinfectant’ logic in those same silent updates that update the code-signing CRLs, same as Microsoft does with Windows Defender. So the usual “join a botnet, hijack your browser” kind of malware can simply be reverted.)

Plus, there’s the whole System Integrity Protection thing, meaning that macOS malware can’t really do anything to permanently subvert the Gatekeeper infrastructure, since it lives in the “untouchable” root partition. (It could do something clever with a system extension, but as of Catalina you have to explicitly activate those in the Security preference pane; and probably, as of Big Sur, you won’t be able to activate them at all.) So it’s only people with SIP off (i.e. system extension developers; Hackintosh owners) who would even feel any sort of “deep impact” from any of this malware. Meaning that macOS malware authors basically don’t bother to try to “deeply embed” their malware into the OS, given that the process will only actually work on a tiny fraction of systems.

Anyway, all that being said: it’s not like Apple said they can’t “guarantee” that the app is free from malware, implying that signed+notarized apps would be guaranteed free from malware. They just say they can’t “validate” that the app is free from malware, implying that the apps that don’t show this warning have been “validated” by Apple—i.e. audited, to the best of their own abilities and current knowledge. Signed off on, like a home inspector signs off on a house. And that’s exactly the case. Apple has “validated” those apps. That doesn’t translate to some technical guarantee of safety, like running the app in a VM would give. It only translates to “you can trust this app to the degree that you trust Apple’s validation process.”

It’s exactly the same claim that Chrome and Edge are implicitly making when you download software through them on Windows: the software gets “validated” by Google/Microsoft as not containing malware to the best of their knowledge. It’s an antivirus signature scan, combined with a trustworthiness heuristic based on whether the developer was willing to sign their software. The only difference is that, in Apple’s case, the “antivirus scan” part happens on a server somewhere, asynchronously, rather than on the client. But it’s the same level of effective security.