It's a little known fact that the Chinese government has perfected a way to grab user data from offline AI systems by stealth. Every morning at 3am, they send a swarm of pigeons to every offline AI user's home, burrow into the PC area and steal the data that has been generated that day. They then fly back to a Chinese submarine off-shore and unload the cargo. It's fiendishly clever, and a great way to overcome OpenAI's lead in LLM technology.
IMO, there's a deep irony here: yes, DeepSeek the service is Chinese, but the model/weights are open, so we can easily run the models on U.S. infra (e.g. how Perplexity) is currently. We can even run the model entirely offline if we want.