> A company needs to make a reasonable effort.
It's not "reasonable" to demand any effort that won't be effective. Period.
> GNU Taler figured out a way to verify age without revealing your identity.
Well, no. GNU Taler proposed a very complicated system on top of some zero knowledge concepts that have been around since the late 1990s, embedded in the rest of Taler's incredibly complicated overall system. That sort of gross impracticality, generally in the service of authoritarian goals that aren't good ideas to begin with, is how Taler rolls, and why Taler has even less chance of ever getting anywhere than cryptocurrency in general.
But it's true that zero knowledge systems can in theory solve most of the problem: whoever's certifying age doesn't know what the user is doing with the credential, and the relying party doesn't know who the user is.
... and that still has no practical impact at all. You can do all kinds of things in zero knowledge, if you have the infrastructure for it. But to make AV ubiquitous, you first have to make that infrastructure ubiquitous.
In practice, that will probably only happen if every government in the world starts issuing zero-knowledge credentials; only governments have the reach and the generally accepted standing to do that kind of thing. But governments in general have never shown any sign of having any enthusiasm for zero knowledge, in part because they tend to like having somebody's identity attached to everything that happens. At most you might be able to drag governments, kicking and screaming, into adopting systems where only they can invade your privacy. But probably not, because they don't trust one another.
There is, of course, a possibility of a private system developing, or of governments getting off their butts. There's even a possibility of Taler's grandiose dreams of massively complicated interacting systems of public and private institutions coming to pass. However, we don't have any of those now, and even if there were the will to make them work, it'd take decades to make them ubiquitous enough.
So people who want to advocate for age verification need to first go off and do about 50 years of work to build that infrastructure, before they can start talking about requiring AV for anything.
In reality, of course, the intent of all real AV advocates is to totally ignore that possibility and rely on "trust me" assurances from AV vendors.