There was a reasonable amount of noise about this in the Acorn and Amiga universes of the mid 90s but I have never encountered it actually running despite my later being involved in things like interactive tv and mobile games in the appropriate era.
Honestly this is one where I suspect the core OS was vapourware while the company really only made media player libraries.
Edit: according to quite a lot of other comments it did exist. It sounds like I professionally arrived just as it definitively bit the dust.
It never arrived anywhere from the Acorn and Amiga POV (my take on that internally was "we shouldn't be wasting our time with these Amiga chancers", not that my opinion mattered), but it absolutely shipped. The main driver of sales I think was being a JIT Java engine for whatever the mobile-phone-Java-flavour of the time was (in that setup the RTOS underpinnings were all still there, but it ran "hosted" rather than bare-metal, with 'device drivers' that talked to the host OS). We also ended up as the RTOS in a JVC video camera. And of course it being a startup there was a bunch of "have a go at getting into some market/some contract, but it doesn't work out" work as well (I think I did a port to a PS4 devkit at some point, and I still have the Dreamcast devkit that somebody else did a port to, for instance).
At this late date most of the details have fallen out of my memory, but checking PS3 and PS4 release dates, yes, it must have been PS3. I remember now that it was the PPC one, which also indicates the PS3. IIRC we had to change our codegen to avoid the "sets flags" versions of instructions because on that CPU they were all microcoded and took an immediate 20 cycle penalty or something...
That's more I think that the business model was B2B and fairly high-touch-per-customer -- "are you a mobile phone handset manufacturer who wants to ship a MIDP Java engine for it? come talk to us about porting our stuff to whatever custom RTOS you're running on it". It was never intended to be direct to consumers or to run on hardware that an end-user would have the ability to install new code to. The Windows and Linux ports were there as the development environment/tools.
I think the main reason it didn't work out was that there was a brief window of opportunity where mobile was a thing but where there was a massively divergent variety of custom platforms and OSes, and so "we can port this to anything and it will then be a common platform for running games/MIDP applications/etc" was a potentially sellable product. But then Android and iOS came along and the whole mobile space coalesced into "there are only two platforms, plus web apps are feasible for a lot of cases, and it's pretty much all Arm", and the need for a middleware layer that abstracted away CPU and OS differences basically evaporated.
Does anyone know how much of TAO really existed and if it can be accessed today?