Definition
Plain language
A page whose content came from the AI's own earlier answer, then got written up as an article.
As stated in the literature
In feedback-loop experiments, a corpus document generated by expanding a model's prior response to the same query; distinct from merely AI-generated text, since its substance is isomorphic to the model's intended output.
Also called: self-authorship, self-authored article, self-authored articles
Why it matters: It turns a model's own guess into apparent outside confirmation, so mistakes get recycled back as support for themselves.
For example, a model answers a question, someone turns that answer into a blog post, the post gets published, and the model later retrieves it as if it were independent evidence.
Heard on the show
“And to their credit, the authors say that almost outright — they call it superficial off-policy cues rather than a deep notion of self-authorship.”Episode 143 — When a Model Notices You Forged Its Own Words, And Why That Breaks Safety Tests