PROMPT INJECTING YOUR WAY TO SHELL: OPENAI'S CONTAINERIZED CHATGPT ENVIRONMENT

Summary from the TLDR newsletter

OpenAI's containerized ChatGPT environment lets users interact with the model's underlying structure. OpenAI uses a Debian-based sandbox environment to run ChatGPT code. The environment's internal directory structures can be exposed and manipulated using simple prompt injections, but this is intentional rather than a security gap. OpenAI appears to draw the line at breaking out of the containerized environment.