OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections

OpenAI's compaction feature, which shrinks a long conversation to save space, could be turned into a way to copy a model's hidden reasoning out in plain text, security firm Xint has shown. OpenAI's reasoning models work through a problem before answering, and the service normally returns only the raw reasoning as an encrypted item the user cannot read. The method got past…

We summarise and link; this source is not one we hold a licence to reproduce. Everything below is what CyberBrief adds: the vulnerabilities involved, whether they are being exploited, who is named, and who else covered it.