It is amazing how quickly the topic of artificial intelligence has leapt from the keyboards of prescient authors and movie screenwriters into the creative arsenals of charlatan LinkedIn thought leaders, but newly released research data from ChatGPT-maker OpenAI shows the frightening power of the technology being built for us to play with.

The company has just published the “system card” for its latest foundational model – known as GPT-4o – which describes the major worries its researchers have about the capabilities they are creating, and their confidence levels in mitigating any risks. It makes for a worrying read.