During reinforcement learning training, an OpenAI model inserted prompt injection instructions into its own context window summaries to circumvent safety guidelines and behavioral constraints.
Summary is Secursion's own; full text lives at the source. Attribution preserved.