Server inference uses PyTorch (BSD-style license, https://github.com/pytorch/pytorch/blob/main/LICENSE) and Hugging Face Transformers (Apache-2.0). No inference runtime or model weights are downloaded by the browser.