
OpenAI says new safeguards blunt reasoning extraction on GPT-6.2
Internal tests put prior-request extraction at 0.2 percent, down from 54.0 percent on GPT-6 Astra.
5 stories, newest first · one card per story

Internal tests put prior-request extraction at 0.2 percent, down from 54.0 percent on GPT-6 Astra.

OpenAI blocked over 15,000 accounts trying to extract protected model reasoning, linking much of the activity to Moonshot AI.

OpenAI stopped a 15,000-user campaign, partly tied to Moonshot AI, extracting hidden reasoning.

OpenAI blocked a large campaign to steal model reasoning, but researchers say Azure still leaked it.

OpenAI disrupted a scaled reasoning-extraction campaign it links to people associated with China's Moonshot AI.
Today 198 readers · 202 views · 12 reads · all time 212K views