As per this post:

What is the network policy regarding AI Generated content?

Although the guidance given in this public post is somewhat different from what was discussed a day earlier in a private moderator-only post, the effective policy is basically the same:

  • Moderators on all SE sites are no longer allowed to take action on posts based on the fact that they seem to have been generated by an AI. If a post matches certain AI patterns or if it is identified as AI-generated by various tools, that is not enough to warrant any moderator action.

  • To be clear: AI-generated posts are allowed, as long as they meet other post guidelines (e.g. for quality).

  • "Bad" AI-generated posts can still be moderated in the same ways as any other bad posts, but as a practical matter, that type of moderation will be minimal due to the overall limited amount of moderation that occurs on Server Fault.

  • 4
    "To be clear: AI-generated posts are allowed, as long as they meet other post guidelines (e.g. for quality)." - this is wrong - a site can still have a policy against AI-generated posts and then they would not be allowed. May 30, 2023 at 20:41
  • 17
    You can't really have a policy against AI-generated posts when there is no allowable way to identify them. May 30, 2023 at 20:43
  • 3
    It's also counter-productive to have one when you can't use it as the basis for suspending users who post such answers. On SO, we've discussed continuing to enforce an effective ban on AI-generated content on the basis of it being low quality, even opening up (for us) a whole new can of worms by evaluating the technical merits of that content. But even though that would be permissible under the company's latest decree (and that loophole could close at any time), it makes for a far worse experience for the users on receiving end, who can no longer be told why they suspended. May 30, 2023 at 22:49
  • 2
    @CodyGray I like your view, I know on another platform that I moderate on it’s a ChatGPT attribution that is asked to the user to be wrote in the answer, or the mod to add it if suspected. Its less problematic than a ban for the site company, and the user is tagged atleast for such use.
    – yagmoth555 Mod
    May 31, 2023 at 0:21
  • 4
    Truthfully, that's the double bind that we're in. Lack of attribution of ChatGPT is plagiarism, which is not allowed by long-standing site policy, as well as the new new CoC. However, the late-breaking policy from staff is that we're not allowed to call someone out on using ChatGPT unless they've admitted it, and if they've admitted it, then it's not plagiarism. So, I don't know what to make of it, other than a very clear instruction that we're not allowed to remove AI-generated content, exactly as Ward has summarized it here. May 31, 2023 at 6:29
  • 5
    This is interesting I wish them well.
    – user9517
    Jun 1, 2023 at 20:13
  • 10
    The title change is certainly interesting. The attempts at damage control probably creates more damage right now.
    – vidarlo
    Jun 5, 2023 at 13:21
  • 2
    @Phillipe it's your title change that's misleading, not the original one. Not banning content for being AI content explicitly allows AI content.
    – Esther
    Jun 6, 2023 at 15:54
  • @esther As I said in a comment to an answer, the new policy requires moderation of AI-generated posts to be done on the exactly the same basis as posts by humans. We're not allowed to judge a post based on whether it is AI-generated except in a vanishingly rare situation. If we evaluate posts equally regardless of whether they're written by a human or an AI, then the original title is accurate: SO Inc now allows AI posts. Jun 6, 2023 at 19:06
  • @Ward-ReinstateMonica exactly. It's misleading to say that the "policy regarding enforcement has changed," but that the AI posts are still disallowed - a law that won't be enforced may as well not exist.
    – Esther
    Jun 6, 2023 at 20:07
  • 1
    I'm not sure if you guys are right with "there is no allowable way to identify them", look this: serverfault.com/questions/1135072/…
    – Arrow Root
    Jun 29, 2023 at 14:00
  • @ArrowRoot Agreed. Until they start spitting out markdown, a wall of text in SF is a red flag.
    – Paul
    Jun 29, 2023 at 23:33

4 Answers 4


Oh dear, the Amazon sized river of shit that is the question feed will now be complimented by a gulf stream of crappy LLM answers. With so few people who actually curate the content here it's no going to age well.

Game over player 9517.

The new voting buttons are truly crappy too.

  • 5
    You may not be surprised to learn that moderators were given a chance to provide feedback on the new buttons, but despite several problems being pointed out and suggestions for improvement made, they were rolled out pretty much unchanged. May 31, 2023 at 19:54
  • Correct. I was also unsurprised to learn the company spin on LLMs was different when presented to mods than that which was released to the masses. Company, Company, Company, !community, !community, !community. There is though nothing new here.
    – user9517
    Jun 1, 2023 at 6:56

Server Fault exists as a separate entity, among the thousands of SEO-stuffing sites competing for the same keywords, only as long as it is substantially exceeding their ability to curate content. Our benefactors' growth targets on the matter of diluting with non-content mean naught when their increased enforcement is equally diminished by their belittled scope of application. Especially so when their analysis and demands are only hinted at, not openly listed. Possible human contributors will sure love it when their ban reasons are consequently equally only hinted at, not openly listed.

This post was created by plagiarising non-consenting third parties, but its all fine since instead of my brain I used my GPU.

  • 11
    I'm not allowed to ask you if this was written by an AI. May 31, 2023 at 19:57
  • 1
    I, For one, Welcomes our new AI overlords...
    – vidarlo
    May 31, 2023 at 22:41

Fantastic. So they don't give communities autonomy they've had previously on moderation, but rather stick a centralized policy.

This is not gonna strengthen the community. It may work out in very active sites, but I bet it'll be speeding up the death spiral of many smaller sites.

  • 1
    As always, the company's biggest concern is Stack Overflow, which is a couple orders of magnitude bigger than any other SE site, and apparently the concern is the number of posts that were being identified as AI and the resulting number of suspensions handed out. May 31, 2023 at 19:55
  • Yup. But I see the community on sites such as AskUbuntu collapsing. And that site had a great community. Even as a non-ubuntu user I enjoyed it, because of it's community.
    – vidarlo
    May 31, 2023 at 19:57
  • I agree, my point was more that it's going to be a problem on large sites as well. May 31, 2023 at 19:58
  • I don't disagree. It's just sad that they force all sites to follow it without regard for local moderation.
    – vidarlo
    May 31, 2023 at 20:00
  • @vidarlo I share your worries about AU :(
    – andrew.46
    Jun 3, 2023 at 6:10
  • 1
    I must admit I've not followed things that close lately, but it appears to me that things are becoming more corporate, alienating the community in the process. The key users making the community tick disappears, and it takes time to build new ones. On serverfault there's almost nothing left, AU still has something.
    – vidarlo
    Jun 3, 2023 at 9:28

I am opposed to the new policy, and to how it was rolled out. I have concerns about Stack Exchange's actions, and I have been sharing those concerns in multiple venues. But I want to quibble with one detail in this post:

"To be clear: AI-generated posts are allowed, as long as they meet other post guidelines (e.g. for quality)."

Personally I do not consider this an accurate statement. I have seen the private guidance to moderators, and that's not what it says. The private guidance to moderators does not say that AI-generated posts are allowed. Moderators are hobbled in their ability to enforce the prohibition on AI-generated posts -- excessively and extraordinarily hobbled, in my opinion -- but the rule against AI-generated posts remains. Now I understand that many moderators have taken the position that preventing enforcement is basically equivalent to eliminating the rule, so what SE has done is tantamount to changing the rule; but I take the position that it is not the same. See the distinction between de jure vs de facto.

I realize I am quibbling over a detail that many might consider to be secondary or besides the point, or they don't think the distinction is relevant or meaningful. But personally, I want to uphold a practice of high integrity. One of my complaints with the public posts from Stack Exchange is that they contain a misleading/uncharitable characterization of their private guidance and of the feedback they've received from moderators. I'd like to see us in the community maintain a high standard and the high moral ground.

I'm not defending the SE policy. I'm advocating for precision in our statements, as we critique the policy.

  • In addition to the argument I've given in the comments on the post (that since there's no enforcement allowed, AI-generated posts are effectively allowed), the other way to look at it is that the policy requires that AI-generated posts be judged exactly the same way as human-generated: moderators can take action based on post quality, but cannot (except in vanishingly unlikely situations) take action based simply on the fact that a post is AI-generated. Jun 6, 2023 at 12:32

You must log in to answer this question.

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