26/09/2026
OpenAI is now publicly documenting when its AI goes wrong — and some of these examples are wild.
The self-replicating prompt injection one especially 👀
Full reports are public on OpenAI’s Alignment site.
Follow for more AI news explained simply.