OpenAI has published a document outlining its multi-layered approach to safety in AI systems. This includes measures to prevent 'prompt engineering' and 'jailbreaks', protect user privacy and security, and details external testing and evaluation processes. The company also notes ongoing efforts to refine these safeguards. --- Why it matters: This matters because it provides insight into how a leading AI research organization approaches safety in its systems, which could inform the development of similar measures by other researchers and engineers. Source: https://openai.com/index/operator-system-card