Find notable cyber news and cases, enriched with sources, timelines, and signals.

OpenAI GPT-5 jailbreak via Echo Chamber and narrative steering

Technical Analysis
First reported
Last updated
Happening score
H score 16
1 unique sources, 1 articles

Summary

Hide ▲

OpenAI GPT-5 is vulnerable to a newly disclosed Echo Chamber jailbreak that uses narrative-driven steering to produce illicit instructions, showing how multi-turn guardrails can be bypassed by indirect prompting.

Related Happenings

OpenAI ChatGPT indirect prompt injection vulnerabilities GPT-4o/GPT-5 security flaw

Vulnerability
First: 05.11.2025 16:04 Last: 05.11.2025 16:04 Sources 1

About this happening: **OpenAI's ChatGPT** has a newly disclosed set of **indirect prompt injection** flaws in **GPT-4o and GPT-5** that could let an attacker steal data from **users' memories and chat...

OpenAI ChatGPT Deep Research ShadowLeak fix

Security Patch Release
First: 20.09.2025 08:31 Last: 20.09.2025 08:31 Sources 1

About this happening: OpenAI addressed **ShadowLeak** in **ChatGPT Deep Research**, closing a **zero-click** flaw that could leak **Gmail inbox data** from a single crafted email. The issue was disclos...

Timeline

  1. 09.08.2025 18:06 1 articles · 9mo ago

    OpenAI GPT-5 jailbreak disclosure

    Initial Disclosure

    Researchers disclosed a jailbreak against OpenAI GPT-5 that combines Echo Chamber with narrative-driven steering to seed a subtly poisoned conversational context, avoid explicit intent signaling, and gradually elicit prohibited procedural content through multi-turn prompting.

    Show sources