Gary Marcus flags OpenAI claiming credit for a known prompt injection attack already cited in its own report

mjdramstead · x · 2026-09-27

Gary Marcus amplified an accusation that OpenAI presented a "new research finding" on self-replicating prompt injection that was actually previously known. Citing @Actuallykeltan: a 2025 experimental demo of the attack is easily searchable, its authors disclosed the findings to OpenAI last year, and that paper is the top citation in OpenAI's own disclosure report — suggesting either deliberate misrepresentation or unchecked AI-generated citations.

Related event: OpenAI Accused of Passing Off Known Prompt Injection Attack as New Finding(3 posts)→

Original post →

More from Fun

Fun channel →