Tell HN: Several ChatGPT Users Report the 'Improve the Model' Training Toggle Turning Itself Back On
Hacker News (459pts, 180 comments)·low signal
User jacquesm wrote that they had logged exactly when they last turned off ChatGPT's allow-training setting and later found it re-enabled. Several commenters described the same thing, while many others said their opt-out had held for months or years. An OpenAI employee replied that an opt-out in either location is respected, but gave no explanation for the reports. The evidence is anecdotal and unconfirmed, and the only practical step is to check Data Controls periodically.