Why AI systems may never be secure, and what to do about it - The Economist

The article identifies that the natural language instruction paradigm of AI chatbots and large language models (LLMs) fundamentally introduces a "systemic ...