In the Prosus-OpenAI deal, what is the situation with deleted answers, please?

Can / will deleted answers be given to OpenAI to use as training?

  • 16
    A) You won't be permitted to delete your answers. SE, the community, and moderators considers deletions and defacing edits to be vandalism, which will be reverted and you, likely, suspended for a moderate period of time. B) It was too late years ago. AI companies already have all the data and have been using it for years. All the user-supplied content is, and has been, available on the web, in the data dumps, in the SE API, and multiple other datasets for years.
    – Makyen
    Commented May 12 at 19:42
  • 3
    Are you going to protest by mass-defacing your posts, getting suspended, and swearing off use of the network?
    – Nij
    Commented May 12 at 22:53
  • 8
    "If I delete my answers from SO say today..." This is not how Internet works. All your data has been fed to the AI long time ago. Even if SO would not be in partnership with AI companies, your data would still be fed to the AI. Even if SO shutdowns today, all your data will still exist in Data dumps and all over the place in various SO copy cat sites. Deletion only hurts people, not AI. Commented May 13 at 6:04
  • 1
    Problem is with the word "my", which implies some sort of ownership. Here is doesn't mean that they are yours anymore, just that you are the creator of them, but that doesn't give you any special rights apart of the right to be named (sometimes, look for the second instance of "perpetual and irrevocable right" in the quote in rene's answer). Commented May 14 at 6:31

3 Answers 3


When you signed up/started contributing you agreed to the Terms Of Service.

It states:

You agree that any and all content, including without limitation any and all text, graphics, logos, tools, photographs, images, illustrations, software or source code, audio and video, animations, and product feedback (collectively, “Content”) that you provide to the public Network (collectively, “Subscriber Content”), is perpetually and irrevocably licensed to Stack Overflow on a worldwide, royalty-free, non-exclusive basis pursuant to Creative Commons licensing terms (CC BY-SA 4.0), and you grant Stack Overflow the perpetual and irrevocable right and license to access, use, process, copy, distribute, export, display and to commercially exploit such Subscriber Content, even if such Subscriber Content has been contributed and subsequently removed by you

(emphasis mine)

So all subscriber content, deleted or not, can, will be and is used by Stack Overflow. It can, under the agreed by you Terms Of Service, distribute, export, display and to commercially exploit such Subscriber Content. It doesn't exclude OpenAI there.

Notice also the word irrevocable. If you (and all other noise providers on social media) didn't want your subscriber content to be used by Stack Overflow as stated in the Terms of Service, you should have signed up to a different site 6 to 8 weeks ago.

I do warn you and future readers that deleting and/or defacing posts is a terms of use violation as well. Y'all don't want to willingly break a contract, right?

  • 13
    Just FYI: The CC BY-SA licenses are, themselves, irrevocable, regardless of if that was reiterated in the Terms of Service. For example in CC BY-SA 4.0 section 2(a)(1) "Subject to the terms and conditions of this Public License, the Licensor hereby grants You a worldwide, royalty-free, non-sublicensable, non-exclusive, irrevocable license to exercise the Licensed Rights in the Licensed Material to..." (emphasis added)
    – Makyen
    Commented May 12 at 19:55
  • "I do warn you and future readers that deleting and/or defacing posts is a terms of use violation as well. Y'all don't want to willingly break a contract, right?" Just for clarity: Deleting and/or defacing posts surely is not a nice action, but also it surely isn't a breach of contract here. It's impossible to delete content on SO, it's always kept in the history. The TOS doesn't govern how people use the interface. Commented May 14 at 6:27

I feel like you misunderstand how deletion works here. It’s soft deletion and it’s all still in the database. We're not sure what's the scope of the APIs available yet, and your posts are probably scraped already. It’s already too late. Your soul is forfeit and your data has been assimilated

I'm also not entirely convinced Prosus directly negotiates these deals. SE is Prosus-owned, sure, but I'd assume that they're legally and functionally distinct.

Let's have a simple straight yes/no answer from Prosus. No endless bumpf or chitty chat chat. "Can / will deleted answers be given to OpenAI to use as training?" - it's a clear up/down question.)

That is never happening.

From a moderator point of view, we're going to treat this the same as any other vandalism, and roll it back. I also feel like we're literally having people on social networks drive outrage for their own purposes.

Basically you're literally causing work to the moderation teams, and possibly the community team as opposed to 'Prosus' or the downstream sites, and deletions don't actually prevent your data from getting fed into datasets used to train AI. There's no sharp cut off. There's also no mechanism for hard deleting an entire post; redactions work partially. Also in this process, your work's lost to the commons, which is a tragedy.

Also, while we're unsure of the implementation details, part of these deals seem to be to ensure attribution, and in theory as long as the CC licence variation our posts are licenced under is maintained. That's to say that posts are attributed and free to share alike, and we can't really tell someone 'we don't like you, go away'.


If I delete my answers from SO say today (Sunday), is it too late for Prosus to give them to OpenAI?

Yes, they will be given to OpenAI to use as training. (Rationale: mods will undelete these deleted answers).

If you only delete very few answers and not too abruptly, maybe mods won't see it. However, even deleted answers could be used in theory by OpenAI (e.g., contrastive learning), though in practice, OpenAI's model designs are not public.

  • 7
    chatgpt infamously had a 2022 cutoff and has been AI model creators voraciously gobbling up everything they could, allegedly running out of data on the internet . Considering SE gets scraped by dubious sites with far less resources, we don't really know if its just mods undeleting, or if they'd scrape alternate sources. In addition, we don't know what the API related to this partnership means. Its worth seperating facts and speculation here, and not encouraging ways to try to get around long standing rules on self vandalism. Commented May 13 at 0:52
  • @JourneymanGeek The most recent GPT models have a cutoff date of December 2023. I've written that even deleted answers may be used by OpenAI, which doesn't strike me as encouraging self-vandalism. Commented May 13 at 17:13

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .