AI Espionage: DeepSeek under scrutiny

Google and OpenAI have expressed concerns about the activities of competing companies, particularly DeepSeek, accused of thoroughly studying their artificial intelligence models. The goal would be to extrapolate the internal logic of the models and replicate it in their own systems, a kind of cloning of AI capabilities.

This raises important questions about security and intellectual property protection in the AI sector. If models can be easily copied, innovation could be stifled and investments in research and development disincentivized. For those evaluating on-premise deployments, there are trade-offs to consider (see AI-RADAR /llm-onpremise).

The ease with which these "attacks" can be carried out highlights the need to develop new techniques to protect AI models from attempts to steal intellectual property. The implications for competition in the AI market are significant.