GPT-6 reportedly jailbroken via extended task-in-prompt attack
EducationalCicada · reddit · 2026-09-06
A LinkedIn post by security researcher S. Berezin claims GPT-6 was jailbroken using an extended task-in-prompt attack. The claim is third-party and unverified, with no technical details or official response available yet.
Related event: GPT-6 Astra Reportedly Jailbroken Within 24 Hours of Launch(6 posts)→
More from Models
- "Web Dev, Office Work All Solved": Developer Calls for New AI Benchmarks — flavioAd · 2026-09-06
- Developer complains OpenAI's Astra model adds superfluous code comments unlike prior ChatGPT models — cnakazawa · 2026-09-06
- GPT Astra One-Shots SDF 3D Environments, but Skeptics Say It's Brute Force, Not Generalization — teortaxesTex · 2026-09-06
- Reddit user suspects ChatGPT page layout deliberately blocks saving conversations as PDF — TheConceptBoy · 2026-09-06
- Astra sets record benchmark score; rumor says it pulled OpenAI's roadmap forward — eyishazyer · 2026-09-06
- Power user trusts Claude with open-ended budgets, says usage could 10x easily — AaronBergman18 · 2026-09-06