I’m not sure if it counts as bug, exactly; it’s not set because Unity runs in batch mode for MR builds, and doesn’t render with the cameras in the scene. Instead, all rendering (to the frame buffer, versus a RenderTexture) is performed by RealityKit and uses its cameras. In bounded mode, there’s no way to know if anything is visible; we don’t even get the device position/orientation. In unbounded mode, we can get the device pose and could conceivably use a conservative estimate of its view frustum to determine what’s visible. If that’s a feature you’d be interested in, feel free to submit it to the road map. Otherwise, you’d have to do the same thing yourself: get the device pose, use it to calculate a view frustum, and test the bounds of the Renderer against the frustum planes.