Social media app Haven will disallow generative AI and de-incentive clickbait, says founder
Kickstarter still going for app developers say won't let AI steal users' biometrics
By the Chilltown Blues
Upload work to a social platform lately?
For some people, the idea of that work being used — sans pay — as part of a pool for generative AI to make a generalized copy is somehow not a pleasant prospect. And as far as pleasant prospects go, opening a social media app may not be one, either, even if someone feels compelled to.
Compulsion in frequent users, according to according to studies by researchers Wendy Wood and Ian Anderson at USC, is the result of the way some of the most popular apps’ underlying mechanisms generally work now, sidelining conscious choice. And when social media is flooded with a rising wave of AI slop, which includes ragebait, that flooding has a tangible effect more difficult for the ave rage person to block out.
William Lavelle is the founder of Haven Social, a social media app its development team intends to make an alternative to “a bleak time to be on the internet,” as their ongoing Kickstarter (through April 10) — https://www.kickstarter.com/projects/havensocial/haven-stopping-ai-art-theft-and-surveillance?ref=dnqcxh — notes.
Whereas other platforms have been making certain relatively new AI features pillars of their applications, Lavelle and the Haven team are developing a social media app where the pillar, they say, is to not do that.
Their other goals include ensuring basic privacy rights for their users and the potential for exploration of difficult concepts to inform, rather than making those concepts reductive for profitability’s sake.
Haven is also a business enterprise where the intention for the business element — advertising that will be atypical, according to Lavelle (more on that below) — is to have an ethical cap.
Haven responded to our questions and thoughts (in bold) on their app below.
Videos on Haven we’ve seen posit Haven’s interface as Tik-Tok-like but pictures and Tik-Tok don’t mix unless they’re in a sideshow, so it seems like format-wise you’re leaning more for Instagram if it leaned a bit more toward YouTube (than it already does post Meta’s catchup) except Tik-Tok has always seemed to prefer shorter-form work. Is that accurate?
Haven: Spot on. In our interviews with VildrArt and Roxane Lapa, we describe Haven as a Instagram/TikTok replacement, because our founder did his Master’s thesis on TikTok. Those interviews were recorded a month before launching the Kickstarter, and once we launched, we received a ton of community feedback about how people want a YouTube replacement instead of a TikTok replacement. So, we’ve shifted to focus on giving Instagram & YouTube some competition.
We’re taking YouTube’s support for long-form landscape video and playlists, Instagram’s support for images and reposting, and combining them in a UI that seamlessly shows both types of content. You’ll be able to easily swap between a discovery algorithm, and a reverse chronological following-only algorithm.
Haven’s intent is to take AI and a lot of invisible malware out of the equation. You’ll be platforming a lot of different kinds of art, music included. If that’s the case, are musicians able to put up covers as they would on TikTok/YouTube -- with the same stipulations inherent to that kind of art?
Haven: For sure! With the flood of AI-generated music onto YouTube and Spotify, Haven will be a great place to listen to music you can be assured isn’t AI-generated. In the future, we’d like to add protections to prevent music from being trained on by AI firms such as Suno, but one thing at a time — we’re cautious to avoid feature creep.
Your YouTube video with VildrArt and inkiio, wherein you speak with people speaking through original art avatars that may appear propped up by a level of AI presentation-wise to someone unfamiliar with that tech, got us thinking.
We’re not for generative AI but don’t mind the “trimmings” necessarily. Like AI voice narrators, for instance. They inherently put someone or a group who doesn’t use a natural, skilled narrator at a disadvantage, and there would be caveats to disbarring that kind of use, which people disabilities. So, how is that line defined on Haven? Is there nuance because the ultimate litmus test is people?
Haven: No worries, VildrArt’s and inkiio’s Vtuber avatars don’t use AI, it’s all deterministic programming. It’s important to remember that “AI” is a massive umbrella term that describes several only barely related technologies, and on top of that, many marketing departments label their product as “AI” even when it’s not, in order to cash in on the buzzword hype. A good example of this is (some) AI narrators. That’s not really generative AI, that’s text-to-speech, or even vocaloid. Both technologies have been around far before 2025’s AI hype bubble.
As far as AI use goes, the line for what is not allowed is “no generative AI” and “generative AI works cannot be the main content.” A Vtuber avatar that uses non-generative AI to make decisions about how the different parts should move is fine. If you’re making a video essay on generative AI, and during it you show some AI-generated images as b-roll, that’s fine, because the images are a small portion of an otherwise human-made video. It’s difficult to lay down a 100% absolute definition here, there will need to be some nuance and human consideration. Thankfully, some (not all) AI image generators include steganography, such as Google’s SynthID, which we can use to objectively identify and remove AI slop when it applies.
“It’s important to remember that ‘AI’ is a massive umbrella term that describes several only barely related technologies, and on top of that, many marketing departments label their product as ‘AI’ even when it’s not, in order to cash in on the buzzword hype.” - Haven
Even before the AI-justified cleansing, STEM jobs have been more of a limited road than has been advertised popularly.
There have been plenty of STEM grads in the U.S., but companies have outsourced some of those jobs — with what we’ve sometimes heard “America doesn’t have enough STEM” grads being part of the reasoning employers tap into sectors of an international workforce they happen to be able to pay less to. Because of this and for other reasons, the field’s work culture, according to one sociologist, can feel toxic.
If Haven is successful, it’s setting up what’s a more viable road for engineers ... not the least because of the above, but also because you’ve also mentioned wanting to fund development of ethical technology. Haven is not the only entity that wants to do that, but it’s worth mentioning because even when we read about something invested/developed by a person/team with an ethical implication, it seems to end up in a system where it’s viewed by a venture capitalist (VC) lens.
Haven: Correct! Thankfully, we have a lot of allies that are also researching privacy protections and ethical tech. We’re looking to minimize the effect of the VC lens by highly prioritizing becoming self-sufficient through ads so further VC funding isn’t needed, and only offering the opportunity to invest to people we know share our values. (anti-AI and anti-enshittification authors, creators, etc.)
If you’re groaning upon hearing that we’ll run ads, hopefully it helps to know that we are doing so in order to stay private and ward off enshittification. In addition, our ads will be better and more ethical than any other platform. To learn what we mean by “ethical,” please check out our interview with VildrArt, here:
Haven plans a button where the viewer of a video can discount their view. Is that finalized? We thought YouTube's rule for a "view" was probably the most "fair" thing about it. Then we looked into it.
According to BuzzVoice, for a view to be notched it has to be watched for 30 seconds if a video is longer than 30 seconds, 90% of the time if it’s shorter; but if ads aren’t watched, neither counts as a view.
Maybe waiting longer for a view to count is reasonable, but the idea of someone being engaged enough to sit through a whole video and then say that experience in its entirety it shouldn't count as a view … it feels like a draw in ways that are both intellectual and potentially as emotional as the way algorithms engage things now.
Haven: Yes, we're pretty set on adding it. However, the details can still be tweaked. If we find that there is unexpected consequences of this, or that people are abusing the feature, we can always adjust it later to achieve the desired result: removing the incentive to ragebait, clickbait, and vagueposting without harming creators genuinely trying their best.
If you like what we’re building, you help make it happen with our Kickstarter. Haven can’t be built without your support. Please check us out here: https://www.kickstarter.com/projects/havensocial/haven-stopping-ai-art-theft-and-surveillance?ref=dnqcxh