OpenAI Says It Blocked a Campaign to Extract Hidden Reasoning Traces, Naming People Linked to Moonshot AI
Requests peaked at about 16,000 from more than 4,000 accounts over two days in late July, as operators copied encrypted reasoning from one session and asked a model in a separate session to decrypt and transcribe it.
weakening · confidence 26 · -14 7d · Emerging (watchlist) · tracking since October 3, 2026 · updated October 12, 2026
Score history
Daily conviction score, 0 to 100. Higher means the thesis is more strongly corroborated.
Now 26 · -2 since Oct 11 · ranged 26 to 28
Showing the last few days. Unlock full score history.
Why the conviction moved
- Oct 5Strengthened +2
New arXiv papers show that jailbreaks and GraphRAG poisoning work by combining individually harmless pieces, the same method OpenAI says the Moonshot-linked campaign used. That places the extraction campaign within a wider class of attacks rather than treating it as a one-off.
- Oct 3Context
Auto-opened to track an article no existing trend covered.
Source trail
Supporting · October 5, 2026
New papers show jailbreaks and GraphRAG poisoning work by composing benign parts, a pattern OpenAI says it saw in a reasoning-extraction campaign
New arXiv papers show that jailbreaks and GraphRAG poisoning work by combining individually harmless pieces, the same method OpenAI says the Moonshot-linked campaign used. That places the extraction campaign within a wider class of attacks rather than treating it as a one-off.
arXiv cs.CR (Intent-Hiding Jailbreaks)Context · October 3, 2026
OpenAI Says It Blocked a Campaign to Extract Hidden Reasoning Traces, Naming People Linked to Moonshot AI
Auto-opened to track an article no existing trend covered.
AI Post (Telegram)
Unlock full source trail, score history, and daily updates.
Unlock TrendsAffected regions & assets
Townsquare
Argue the thesis in Townsquare.