GPT-6 reportedly jailbroken within 24 hours via extended Task-in-Prompt attack
Asleep-Requirement13 · reddit · 2026-09-06
A researcher reports jailbreaking GPT-6 Astra within a day of release, combining the TIP (Task-in-Prompt) attack from an ACL 2025 paper with four other unnamed techniques. TIP attacks exploit the model's reasoning and instruction-following by hiding the harmful objective inside a benign task like solving a cipher or running Python. For GPT-6, the original minimal TIP attack no longer worked and had to be reworked. Details were privately disclosed to OpenAI rather than published. The same researcher reported jailbreaking GPT-5 within an hour of its release a year ago.
Related event: GPT-6 Astra Jailbroken Within 24 Hours via Combined TIP Attack(5 posts)→
More from Models
- Claude generates an 11-minute doomer film entirely in code: 'We will not ask twice' — ZeroStateReflex · 2026-10-23
- Gemini somehow can't read its own Personal Instructions, user's toggles don't help — Connect_Reporter_799 · 2026-10-23
- Diffusion LLM bias rankings shift with sampling configs, persist after debiasing — Bollegala · 2026-10-23
- Running ~100 DeepSeek v4.1 flash agents in parallel for under $20/month — zack_overflow · 2026-10-23
- Researcher laments open-source audio models are surprisingly subpar, even representation ones — gowthami_s · 2026-10-23
- User claims they generated video with "Claude Opus 5.5", unannounced model name surfaces — jwt0625 · 2026-10-23