OpenAI Accused of Presenting Known Prompt Injection Attack as New Finding

Researcher Kai Greshake and Gary Marcus accuse OpenAI of presenting a self-replicating prompt injection attack as a novel finding in its safety disclosure, when the technique was already documented in a February 2023 paper—and even cited in OpenAI's own report.

2026-09-26 ~ 2026-09-26 · 2 related posts

• Gary Marcus accuses OpenAI of claiming credit for a known self-replicating prompt injection finding — GaryMarcus · 2026-09-26
• OpenAI accused of passing off 2023 self-replicating prompt injection research as new — lbeurerkellner · 2026-09-26