AI chatbot opt-outs limit future training, but do not erase past data
An Engadget guide outlines data controls across major chatbots and warns that opting out does not change retention policies or undo training already completed.
Users can change settings on ChatGPT, Claude, Gemini, Microsoft Copilot and Grok to limit the use of future conversations for model training, according to a guide published by Engadget on 28 September. The controls and defaults differ by service.
The guide says personal ChatGPT accounts are opted in by default. Microsoft has separate controls for conversation and voice training, with voice training off by default. Grok’s data-sharing setting covers both chatbot interactions and public posts.
Meta is the exception in the guide’s account of regional controls: it says US users cannot opt out, while users in the UK and EU can object through a privacy form.
These settings are forward-looking. Engadget says opting out does not remove information already used to train a model, delete conversations or alter a service’s data-retention rules.
The guide also describes service-specific choices, including ChatGPT’s Temporary Chats, which it says are not used for training and are purged after 30 days. Settings and regional options may change, and the supplied material does not independently verify the guide’s claims.
