Researcher Finds Loophole in OpenAI's Training Opt-Out Terms

Researcher Shashwat Goel found a possible loophole in OpenAI's consumer opt-out terms: hidden chain-of-thought may not count as 'Output,' so opting out of training might not cover hidden CoT data.

2026-09-11 ~ 2026-09-11 · 2 related posts

• OpenAI opt-out terms may have a loophole: hidden CoT likely doesn't count as 'Output' — dhruv2038 · 2026-09-11
• OpenAI's consumer opt-out wording may not cover hidden CoT, researcher warns — scaling01 · 2026-09-11