Since the operator uses user's computer via screenshots and keyboard/mouse, it's not easy to block by IP, browser fingerprinting, or convoluted page markup. The initial implementation probably has some detectable traits in its mouse and keyboard use, but a company that could train pretty convincing text and voice models can easily train a model to move mouse naturally enough. Simple screenshots can be messed with via flicker (using persistence of vision), but that's fixable too.
This cat and mouse game can end up ugly.
My pessimistic prediction is that Google et al. will extend their DRM from video to entire Web pages, and sell it as a service that blocks "unauthorized" screenshots and soft mouse, along with a cryptographically-strong way to detect the closed-source Chrome. It won't be accessible, but conveniently Republicans have added accessibility to their list of woke things to destroy. Sites will submit to this and advertise it as "use Chrome to skip those horrible CAPTCHAs!"
This cat and mouse game can end up ugly.
My pessimistic prediction is that Google et al. will extend their DRM from video to entire Web pages, and sell it as a service that blocks "unauthorized" screenshots and soft mouse, along with a cryptographically-strong way to detect the closed-source Chrome. It won't be accessible, but conveniently Republicans have added accessibility to their list of woke things to destroy. Sites will submit to this and advertise it as "use Chrome to skip those horrible CAPTCHAs!"