GenAI is a Privacy Nightmare and LinkedIn is Sneakily Stealing Your Content
Here's what you need to know to protect yourself from a breed of dark patterns
You’re all opted in to something dodgy, here’s how you opt out
🥂 Congratulations! If you’re a LinkedIn user outside the EU, you’ve just been sneakily opted in… to letting LinkedIn clone your posts without crediting you.
Don’t like the sound of that? Here’s the one-click opt-out link: linkedin.com/mypreferences/d/settings/data-for-ai-improvement
Let’s call these opted-in-without-asking shenanigans LinkedIn OptOut for short. In my community on LinkedIn, most people are opting out as quickly as their fingers will let them, while a smattering is shrugging its indifference, and a few folks are staying opted in on purpose. One of you in this last group said you don’t want AI to be trained only on all the “drunk uncles” out there, so you are doing your bit for LinkedIn staying classy. 🤔
So let’s unpack what’s going on with LinkedIn OptOut and why it matters.
What happened?
Spoiler alert: I’ve put a few bangs (factorials!) in the next paragraph to highlight some of the salient points we’ll be chewing on shortly.
Thurs, Sep 18, 2024
If you had a (non-EU!) LinkedIn account then you were automatically opted in to allowing LinkedIn (and its affiliates!) to use your intellectual property to train their automated content creation GenAI systems. LinkedIn provided an opt-out button (silently!) which you’d have found out about if rumbles of discontent had made it to your feed — hat tip to Ben Maling for bringing my attention to it. A look at the fine print shows that opting out does not cover your standard AI for user experience optimization. (It’s all about content creation!)
EUsers
Because the universe loves its irony, this happened on the same day that the UN published its major report on global AI regulation. (Report info and link in the footnotes* if you’re interested.) Their TL;DR? National governments can regulate all they want, but AI doesn’t respect borders.
One thing that did respect borders that day was LinkedIn’s OptOut. All the EU saw was us elsewhere-based users whingeing and sharing a broken link to nowhere.
Personally, this marks the first time I’ve felt like my online experience would have been improved by being in the EU. A new challenger enters the arena!
(The UK was also sneakily opted in until they pushed back, rejoining the EU… in this one small way, at least.)
Defaulting in
What’s the big deal, anyway? Is it that yet another company is training their AI systems on user data. That’s old news. After 10 years of working on AI at Google, it’s just about the most quotidian thing you could tell me.
So is it that LinkedIn opted us all in without asking us first? That’s what a lot of folks seem to be getting vocally upset about today, but I think that misses the heart of the issue. I’m not surprised when users are opted in to online settings by default. Doing it that way is — dare I say it? — a digital best practice for user experience (UX) design. If you had to spend your day giving permission for every bit of code you’re interacting with on the internet, I’m not sure if you’d have achieved logging into your email inbox by bedtime.
As icky as that approach to consent might feel in other contexts, plenty of the GDPR-bashing we’ve seen over the years has been about how constant nagging for user consent has ruined the European internet experience (and then everyone else’s too, since investing in building a non-EU version doesn’t always suit the corporate wallet). When it comes to cookies, most users don’t bother with all the extra steps it takes to select all the rejection options. Time is money and there are cat videos to watch!
And as for legalese, well, companies know they can always annoy a user into giving instant “informed consent” by presenting far too much fine print for any normal person to be bothered with.** (Have you ever read every word of a website’s Terms and Conditions? I have. Once. It’s booooring.)
No, the silent nature of the global opt-in is not the point either. Let’s not get ourselves too emotional about it if we’re guilty of years of smashing ACCEPT ALL without reading. Self-consistency is attractive.
The much more important issue is what we were opted in to… and that it’s quietly against our best interests. This is a dark pattern: the default setting does not match what most users would choose for themselves if asked. You can default opt me in to receiving a glass of clean drinking water with every meal, but please default opt me out of putting vodka in it.
Opted out of opting out
Is the problem that LinkedIn is training AI systems on your data? Nope. That’s not news, it’s business as usual for the internet.
Speaking of cookies, you’re probably pretty used to your data being used to “optimize your online experience” (which means you might get a different experience in incognito mode or on a buddy’s device).
So let’s look at the fine print, shall we?
The opt-in/opt-out button does not apply to the “development of AI models used for other purposes, such as models used to personalize your LinkedIn experience.” Ahem, excuse me?
Turns out LinkedIn OptOut (and this discussion) is not about the kind of data usage by companies that you’ve grown accustomed to. LinkedIn will continue training its standard AI systems on your data and you can opt out with your feet if you don’t like it. This is about something else.
But what? To find out what the real issue is, read on here. We’ll need to make some choices as a society, so we need to build our literacy on this topic before our choices are made for us.
Thanks for reading! How about a course?
If you had fun here and you’re looking for an unboring course designed to delight AI beginners and experts alike, here’s a little something I made for you:
Or my new leadership-oriented enterprise-focused (but still fun) AI course:
Prefer to hone your decision skills instead of building your AI muscles? You can learn decision intelligence from me via this link to my free course:
The steering wheel for your life — Decision Intelligence Video Tutorial | LinkedIn Learning…
Decision-making is the most valuable skill you can learn. Your life boils down to two things: the quality of your…bit.ly
P.S. Have you ever tried hitting the clap button here on Medium more than once to see what happens? ❤️
Liked the author? Connect with Cassie Kozyrkov
Let’s be friends! You can find me on Twitter, YouTube, Substack, and LinkedIn. Interested in having me speak at your event? Use this form to get in touch.
Footnotes
*The United Nations Secretary-General’s advisory body on AI published a report yesterday arguing that AI regulation at the local- and national-level is problematic and therefore a global regulatory framework for AI is essential. Arguments include the rapid fragmentation of standards related to AI and the potential interoperability issues. (I’ll note that standardization and interoperability issues tend to be a pretty good target for AI to help with, in the same way that AI is being used successfully to attack technical debt today.) Here’s the PDF: un.org/sites/un2.un.org/files/ governing_ai_for_humanity_final_report_en.pdf
If you’re keen to see the discussion about whether global AI regulation is practical, here’s the link to my post.
**Since computers are much better at memory and attention than humans (that’s the point of data!), getting some AI assistance with long and boring documents is one of the most obvious personal use cases for GenAI. Having started a company fairly recently, I’m still in the phase where — despite having an attorney — I still check every single contract myself but I’ve been able to do this an order of magnitude faster with some carefully crafted prompts into what I’ve taken to lovingly calling ChatGPT Esquire. Do I do this on full auto? Of course not. I use AI as a tool for speeding myself up, not for removing my brain from the task. I still check all the important bits with my standard issue human eyeballs.
*** This comment suggests that while you might have had the opportunity to change your opt-in/out settings a few weeks ago, your opt-out would have been reversed yesterday. If true, it points to an even darker design pattern, since we tend not to expect to have to go and re-opt-out of something we’ve already put the effort into saying no to.