from Hacker News

LinkedIn will soon train AI models with data from European users

by skilled on 9/22/25, 9:33 AM with 83 comments

  • by rckt on 9/22/25, 10:34 AM

  • by OtherShrezzing on 9/22/25, 10:45 AM

    As far as I can tell, LinkedIn's content is already 99% LLM generated posts.

    The resulting models might be a terrible hybrid distillation of GPT5 and Claude with a strong preference for hustle culture & banal parables.

  • by elzbardico on 9/22/25, 10:23 AM

    So, soon we will see AI generated job descriptions for American companies offering 6-month paternity leave and 30 days/year paid vacations?

  • by zelphirkalt on 9/22/25, 10:54 AM

    I've said this elsewhere: The EU should crack down on this. Opt-out is not acceptable behavior.

  • by thimabi on 9/22/25, 10:44 AM

    These days LinkedIn is a cesspool of AI-generated feel-good “inspirational” content anyway… I don’t see anything good coming out of training models from its user data.

  • by Havoc on 9/22/25, 11:54 AM

    I wish them the best of luck with that.

    Short of 4chan it's hard to think of a worse dataset. Do you really want a model trained on corporate thought leader gibberish?

  • by heikkilevanto on 9/22/25, 10:23 AM

    Glad I never had an account on LinkedIn

  • by blitzar on 9/22/25, 11:03 AM

    This is an incredibly dangerous development - when Ai learns the grind and hustle it is going to accelerate the timeline to Judgement day.

    Only the people with the right grindset that do a 4am ice face bath will survive.

  • by senko on 9/22/25, 10:35 AM

    > The company says it will rely on “legitimate interests” as its legal basis and will offer an opt-out so members can refuse use of their data for training

    "Legitimate interest" is a very specific term in context of GDPR. Not a lawyer, but have been looking into it previously, and I doubt "we want to feed data to our AI so we can make more money" passes the Legitimate Interest Assesment (LIA) test.

    Here's an example of a test that must pass (sorry, docx, but way better than a random explainer): https://ico.org.uk/media2/for-organisations/forms/2258435/gd...

  • by untrimmed on 9/22/25, 11:10 AM

    The part that really gets me is that opting out doesn't affect models already trained on my data. It kinda feels like closing the barn door after the horse has already bolted.

  • by toolis on 9/22/25, 10:56 AM

    i'm excited to find out what AI will learn about b2b sales from all the cringe posts.

  • by bananapub on 9/22/25, 10:34 AM

  • by tracker1 on 9/22/25, 6:27 PM

    What do you expect from a company that bypasses security models on mobile devices to intercept emails? Of course it's opt-out... I'm guessing US users don't get any option.

  • by drsalt on 9/22/25, 10:20 AM

    good insight

  • by supermatt on 9/22/25, 11:08 AM

    What is it with American big tech and blatant disregard for EU privacy laws.

    “Legitimate interest” is being used here to skirt around the need for consent (I.e. explicit opt-in) specifically because they know users will not give consent (otherwise they could easily ask them). That is an immediate red flag when it comes to “balancing” the rights and needs of the user vs the data processors “legitimate interest”.

    i.e. big fine incoming for LinkedIn.

  • by rich_sasha on 9/22/25, 11:09 AM

    Quality of LinkedIn feeds is the lowest of all social network sites I occasionally view. Instagram at least has attractive people and puppies.

    I don't know if it's AI slop or people genuinely baring their souls and revealing corpo-BS pseudo-deep thoughts, with a drop of grifters pushing their BS "Uber but for nailcare on Blockchain" whitepapers, either way I don't see any value you could extract from it.

  • by 0cf8612b2e1e on 9/22/25, 3:23 PM

    I am flabbergasted that Microsoft thinks they need just a more data. Surely we are way past diminishing returns on what some phony, inspirational work memes can deliver.

    Why take the legal risk?

  • by Tubelord on 9/22/25, 10:35 PM

    Training models on so much AI-generated content will certainly lead to diminishing quality.

  • by intellectronica on 9/22/25, 12:38 PM

    Maybe it will finally start recommending better cheese. My feed is so boring.

  • by guywhocodes on 9/22/25, 10:32 AM

    I wonder what this will do for enterprise sales posting

  • by thiht on 9/22/25, 8:07 PM

    Why the hell does LinkedIn need to train AI models?

  • by thatjoeoverthr on 9/22/25, 10:47 AM

    There's a film about this, Multiplicity.

  • by numericcitizen on 9/22/25, 11:08 AM

    LinkedIn is becoming the other Facebook...

  • by tropicalfruit on 9/22/25, 11:28 AM

    is this what they mean by eat your own dog food

  • by siva7 on 9/22/25, 10:45 AM

    Great! Now AI will be optimized for humble-bragging about thought leadership and congratulating connections on work anniversaries they forgot about.

  • by gethly on 9/22/25, 11:58 AM

    It is infuriating that this thing is opt-out instead of opt-in, irregardless of advanced notice to configure it.

  • by sgt on 9/22/25, 11:01 AM

    Seeing that LinkedIn these days is 90% AI-slop, wouldn't this mean that the AI is being trained by AI?

  • by alex1138 on 9/22/25, 10:56 AM

    Because of course they do lol

    Stop AI-ifying your platforms