X Training Grok AI On Tweets Without Notifying Users

[MacRumors]

Social network X (formerly Twitter) recently activated a setting that gives it permission to train Grok AI on user tweets. All X users are opted in by default, with X failing to notify customers about the change.

The hidden setting gives X permission to use all posts, interactions, inputs, and results for "training and fine-turning" Elon Musk's Grok AI model.

To continuously improve your experience, we may utilize your X posts as well as your user interactions, inputs and results with Grok for training and fine-tuning purposes. This also means that your interactions, inputs, and results may also be shared with our service provider xAI for these purposes. Learn more

While X turned the setting on for all users by default, it can be turned off on the website. Here's how:

1. Open up X using a web browser on a Mac or PC and log in.
2. Click on More.
3. Choose Settings and privacy.
4. Click on Privacy and safety.
5. Click on Grok.
6. Uncheck the setting that allows Grok to use your content.

This privacy setting is not available through the X app, so it must be disabled using the desktop version of X, though an option to turn it off on mobile devices is supposedly coming soon.

https://twitter.com/x/status/1816874083721838921

Turning on a data collection feature surreptitiously and automatically opting in users without their consent is underhanded, and many X users are not happy with the social network's decision.

Sometime in the last month, X updated its Help Center to change the wording around Grok. It previously said that "user interactions, inputs, and results" would be used for Grok training, but now the wording also includes X posts.

June 27 wording:

To continually improve the AI system and provide a better user experience, we may utilize user interactions, inputs and results for training and fine-tuning purposes. This means that when you interact with Grok, your interactions, inputs and results may be used to train and enhance the system's performance.

Today's wording:

To continuously improve your experience, we may utilize your X posts as well as your user interactions, inputs and results with Grok for training and fine-tuning purposes. This also means that when you interact with Grok, your interactions, inputs and results may be used to train and enhance the system's performance.

Grok has previously been able to access "real-time public X posts" to respond to user queries, but it is not clear if X posts were being used for training. The opt-out toggle seems to have been enabled around May, so it's possible that's when user posts were first used for improving the chatbot.

X says that user interactions with the chatbot and X posts are used for the following:

• Enhance Grok's understanding of human language and communication.
• Improve Grok's ability to provide accurate, relevant, and engaging responses.
• Develop Grok's sense of humor and wit to make interactions more enjoyable.
• Ensure Grok remains politically unbiased and provides balanced answers.

Elon Musk claimed earlier this week that Grok was being trained on "the most powerful AI training cluster in the world." Grok is currently available to X premium subscribers.

Article Link: X Training Grok AI On Tweets Without Notifying Users

 

[breenmask]

data from people wanting to shame Elon for this will be sucked into OpenAI and later be used on Apple Intelligence without the users permission lol

 

[DocMultimedia]

Lol. AI trained by Russian bots.

 

[Analog Kid]

I find it hard to get up in arms about this... I don't use Twitter, but if I did I'd just assume the company used my data for all manner of things.

 

[coffeemilktea]

I think that's pretty awesome... my cheesy jokes and goofy takes will live on forever in the AI. 😎

 

[Xiao_Xi]

To the anti-EU and anti-GDPR:

Elon Musk’s X under fire over harvesting users’ data to train AI chatbot

The platform’s lead EU regulator is looking into the matter, which may infringe upon the bloc’s privacy rules.

www.politico.eu

 

[jz0309]

is anyone really surprised by this?

 

[knoxer]

I'd be hard pressed to come up with a worse data set to train an AI with.

 

[vjl323]

Been a Twitter user since Day 4. Once Mr. Musk took over, my use of the site went down a ton [and basically zero after he got rid of 3rd party apps]. It's sad the social media network is still used by others with announcements and such, as it *was* a great way to get info from official sources about things [traffic, weather, alerts from local police, etc]. But now the amount of disinformation on the site, and just incredibly racist posts is mind-blowing. It's sad to see a site that was once just used by Mac and iPhone developers as a way to communicate, become what it has become today. I don't think the founders really had a good idea on how to monetize it, and that's a shame as it meant it could be sold to someone who just wanted to toot their own horn even louder. And now tweets going back to March 2006 are being used to train his AI. All without alerting users to this privacy-invading tactic. Sigh.

 

[seek3r]

I'd be hard pressed to come up with a worse data set to train an AI with.

I'm remembering when a decade ago microsoft's chat bot became super racist after interacting with folks on a much more moderated twitter than x has become, unless they're doing some serious backfill Grok is going to be a racist, antivax, misogynistic cesspool of an LLM.

 

[Analog Kid]

I'm remembering when a decade ago microsoft's chat bot became super racist after interacting with folks on a much more moderated twitter than x has become, unless they're doing some serious backfill Grok is going to be a racist, antivax, misogynistic cesspool of an LLM.

Well, you know the old computer science adage: stupid in, stupid out

 

[swm]

this will be a great addition to the full self-driving semi-sentient Optimus robots that will come sometimes next year in robotaxies

 

[ThunderSkunk]

Littlesnitch, new rule, block all connections to & from twitter.com x.com etc...

easily solved

 

[tubedogg]

Turning on a data collection feature surreptitiously and automatically opting in users without their consent is underhanded

You're willing to claw at Twitter (I refuse to call it X) over this but you seemingly take every other company at face value when parroting their press releases?

Anyway, the issue here isn't privacy (they are public tweets, after all) but copyright. I also don't see how it's "data collection" when the company in question is the one that you submitted the tweet to in the first place.

 

[jedimasterkyle]

If there was ever a point in time when the universe told me to delete my account from something, this is it...

 

[spazzcat]

Lol. AI trained by Russian bots.

Bots training bots ...

 

[tgwaste]

the tweets are already public though..

 

[Lepton]

It’s a fine option. But this whole thing bothers me exactly zero. I can browse websites, media, streams, all sorts of publicly available stuff. And then, I can create my own legal content using what I’ve learned. If I can do that, why can’t a computer? Are you jelly it can read faster and remember better than you? So what? Isn’t the whole point of a tweet to get your words out there and absorbed?

 

[neliason]

Posts on X are public. I’m not sure why anyone would get upset about this default setting. It’s not DMs or anything private.

 

[andrewxgx]

this is why EU has beef with AI companies btw
and why companies have beef with EU, as they can't do what the hell they want there

 

[PBG4 Dude]

I'd be hard pressed to come up with a worse data set to train an AI with.

Truth Social, Gab, VOAT, Kiwifarms, 4Chan…

 

[ninetyCent]

Wait till you see what apple trains their AI on and who they notified.

 

[guerro]

Social network X (formerly Twitter) recently activated a setting that gives it permission to train Grok AI on user tweets. All X users are opted in by default, with X failing to notify customers about the change.

The hidden setting gives X permission to use all posts, interactions, inputs, and results for "training and fine-turning" Elon Musk's Grok AI model.

While X turned the setting on for all users by default, it can be turned off on the website. Here's how:

1. Open up X using a web browser on a Mac or PC and log in.
   
2. Click on More.
   
3. Choose Settings and privacy.
   
4. Click on Privacy and safety.
   
5. Click on Grok.
   
6. Uncheck the setting that allows Grok to use your content.

This privacy setting is not available through the X app, so it must be disabled using the desktop version of X, though an option to turn it off on mobile devices is supposedly coming soon.

https://twitter.com/x/status/1816874083721838921

Turning on a data collection feature surreptitiously and automatically opting in users without their consent is underhanded, and many X users are not happy with the social network's decision.

Sometime in the last month, X updated its Help Center to change the wording around Grok. It previously said that "user interactions, inputs, and results" would be used for Grok training, but now the wording also includes X posts.

June 27 wording:
Today's wording:Grok has previously been able to access "real-time public X posts" to respond to user queries, but it is not clear if X posts were being used for training. The opt-out toggle seems to have been enabled around May, so it's possible that's when user posts were first used for improving the chatbot.

X says that user interactions with the chatbot and X posts are used for the following:

• Enhance Grok's understanding of human language and communication.
  
• Improve Grok's ability to provide accurate, relevant, and engaging responses.
  
• Develop Grok's sense of humor and wit to make interactions more enjoyable.
  
• Ensure Grok remains politically unbiased and provides balanced answers.

Elon Musk claimed earlier this week that Grok was being trained on "the most powerful AI training cluster in the world." Grok is currently available to X premium subscribers.

Article Link: X Training Grok AI On Tweets Without Notifying Users

Now do YouTube, Apple, and META.

 

[contacos]

Is this enabled in the EU? Seems like a huge breach

 

[sfrsteel]

LOL @ the people that are mad at Elon but not Sundar Pachai , Mark Zuckerberg, Jeff Bezos, Bill Gates and Steve Apple.

Social media companies have been mining your data since MySpace... But Your mad at Elon.
FB literally knows when you are going to the bathroom.. But you are mad at Elon..
You got 10 Google Home Mini, 2 Amazon Echo Shows, and 5 Apple HomePods littered throughout your house and work
AND YOUR MAD AT ELON?

The Amount of data that big tech has on you.. They know more about you than you do.. They know your location 24/7 they know what your going to buy, how you are feeling, when you are doing the nasty.. they know everything.

You gladly give away this info every day.
And they sell it to everyone. You are the product..