The field of language model services is moving towards more secure and scalable solutions. Researchers are exploring ways to improve the privacy and efficiency of large language models, including the use of decentralized overlay networks and covert prompt transmission. Another area of focus is the development of distributed retrieval-augmented generation frameworks, which aim to address factual deficiencies and hallucinations in language models while maintaining data privacy. Notable papers in this area include: PingPong, which introduces a new end-to-end system for metadata-private messaging that overcomes the limitations of traditional systems. GenTorrent, which proposes a decentralized overlay network for scaling large language model serving. Covert Prompt Transmission, which investigates covert prompt transmission for secure and efficient large language model services over wireless networks. Distributed Retrieval-Augmented Generation, which introduces a novel framework for improving data privacy in retrieval-augmented generation.