Ollm AI demonstrates V2 proofs

- Ollm AI released verifiable V2 proofs that claim hundreds of zero‑data‑retention local models and TEE confidential models for on‑device use. (x.com) - The V2 bundle specifically names a Dolphin Mistral 24B proof and says device models can run without uploading user screenshots or logs. (x.com) - That pushes the privacy argument for local LLMs while raising product questions about update, auditability, and vendor lock for device‑only models. (x.com)