Apparently the error shows up on the logs if you have a floor in the scene with reflectivity 0. Bumping it to something as small as 0.01 prevents the bogus error from being logged.

Apparently this was a code signing problem. Things weren't exactly matching between app, framework and service, and that caused the failure.

Hi, To avoid installing the beta as the base OS on my work machine, I installed it on a virtual machine running on Virtual Buddy (itself running on Sequoia). However, it doesn't seem appear to be possible to turn on Apple Intelligence on Tahoe running on a VM, and the Foundation Models availability reports as .deviceNotEligible. It does report as .available when running on the iOS simulator inside the VM, but I get the same error as the original poster. Is there any way to use Apple Intelligence, and therefore the new Foundation Models, on Tahoe running on a VM? Thank you, Miguel Arroz