A Tell HN thread reports OpenAI's allow-training setting turning itself back on

Sources: discussion
Summary: An HN user reports that OpenAI keeps re-enabling the setting that allows training on their conversations, and advises checking it rather than assuming it stayed off. No primary source or vendor statement exists for the claim.
Comments: One commenter reports the toggle writes a local storage entry whose value does not appear to affect new tab loads, and suggests reverse engineering the page to tell a UI bug from a reset. One reports not seeing the setting flip on their own account. Another argues that settings resetting across updates is common enough that intent cannot be inferred from the behavior alone, and another notes that documenting the resets would undercut any legal weight the checkbox carries.
Why it matters: The thread cannot separate a UI bug from a default reset, which is the state of the user-facing evidence about what a vendor does with conversations.