OpenAI accused of passing off 2023 self-replicating prompt injection research as new
lbeurerkellner · x · 2026-09-26
- Researcher Kai Greshake notes OpenAI's disclosure report presented "self-replicating prompts" as a novel finding, but his team demonstrated it in Feb 2023 in the paper "Not what you've signed up for" (arXiv:2302.12173) on indirect prompt injection.
- @Actuallykeltan adds the 2023 paper is the top citation in OpenAI's own disclosure, the authors disclosed findings to OpenAI a year ago, and OpenAI either misrepresented it as their own or relied on unchecked GPT-generated citations.
- The original paper built a security taxonomy of indirect prompt injection—data theft, worming, information ecosystem contamination—and proved viability against real systems including Bing's GPT-4 chat.
Related event: OpenAI Accused of Passing Off Known Prompt Injection Attack as New Finding(3 posts)→
More from Safety
- Polymarket odds: 8% chance Anthropic fully pauses AI training by end of year — Polymarket · 2026-11-11
- User fakes hostage crisis to trick Google AI into recommending cracked-software sites — SumitGup · 2026-11-11
- OpenAI DevDay 2026: Codex Security goes from vulnerability to verified fix — OpenAI · 2026-11-11
- Inside the 'abliterators': the renegades stripping guardrails from open-weight AI models — willknight · 2026-11-11
- VulnCheck audit: Anthropic's Glasswing logged 442 CVEs but only 217 fixes in 7 months — jedisct1 · 2026-11-11
- UC Berkeley's SubWM watermark detects LLM distillation even after laundering attacks — sumanthd17 · 2026-11-11