X blames users for Grok-generated CSAM; no fixes announced
- Thread starter JournalBot
- Start date
I’ve not used X since long before musk bought it, but even if Grok content is just suggested for the user to post that doesn’t count for s230 protection: the operator can’t contribute too much to the content. The suggested reply feature is already skating on thin ice.
Upvote
2
(2
/
0)
I don't usually hold Steve Jobs as the genius everyone else does, but one thing to consider, if he was still alive and CEO of Apple, I think he would be crazy enough to give a middle finger to the Supreme Leader allies and block X from the App Store.
Now Tim Cook? Well, he would probably invite them for a walk in Apple Park and give them some cookies
Now Tim Cook? Well, he would probably invite them for a walk in Apple Park and give them some cookies
Upvote
8
(9
/
-1)
Probably more correct that he considers it acceptable collateral damage for generously allowing childish pranks and adult/sexualizing themes. Censoring anything that can approach sexualizing minors, inevitably produces tons and tons of false positives. It's much worse than censoring content produced by humans, because of the scale and because of the lack of real judgement at creation time. A great great lot of innucuous prompts get refused by other services for this reason, creating a "nanny" vibe.
The attitude is actually quite similar to the anti-mask anti-vax stance that the vulnerable people should just stay inside if they want to be safe, so THEY can be FREE.
Upvote
2
(2
/
0)
The Overton Window is being moved further and further by MAGA. Domestic fascism, foreign unbridled imperialism, legitimising CSAM, conspiracy theories, deligitimising science...
Upvote
10
(10
/
0)
I don't think it's being moved nearly as far as the all-online crowd believes.
But I also believe that consensus seeking is pointless navel gazing (when it comes to opposition against Musk and the GOP) and the only way out will be to find a way to cause a crisis that everyone will be forced to react to by rowing in the same direction abruptly, which would invariably result in looting the rich.
It's hard to conceive of a crisis to force that would result in the dismantling of the GOP as a political party and 73-90% tax rates on the GOP while forcing absurd hype-generating stock market valuations to be censored or edited that don't match what a company has on hand or can produce.
But I'm sure that crisis exists and it will probably entail united external pressure from Canada, Mexico, the EU and Japan in response to moves on Greenland or something similar.
Moreover such a crisis that forces everyone to torch the Christian Nationalist cultural mythology wholesale will probably be started by someone that today is all aboard the MAGA train but finally has doubts while in power and takes the role of Brutus to the institutional Caesar of the GOP's finance and donor networks, well after the GOP thinks they've purged any ideological opposition in their inner orbits.
I don't see election or political pressure from the outside alone producing this 'fuck the consensus, fuck the optics, fuck the US Constitution in order to save it, full speed ahead' crisis, it's gonna be someone that currently thinks they have no daylight with Trump and reconsiders, e.g. the more Marjorie Taylor Greenes he creates the more screwed he, Vance, Elon, Thiel and Stephen Miller are.
Upvote
5
(5
/
0)
DogeDesigner is the biggest and most shameless Elon Musk fanboy/gal.
Musk quotes and engages with them so frequently I've often wondered if they are run by Musk..
Their simplistic pen analogy sucks too. I expect more fines on X by the EU.
Musk quotes and engages with them so frequently I've often wondered if they are run by Musk..
Their simplistic pen analogy sucks too. I expect more fines on X by the EU.
Upvote
10
(10
/
0)
Poor Grok seems to be conflicted between its incessant, nauseating desire to please (so that it attracts more X Premium subs), even with illegal outputs, and declining such requests to avoid getting xAI in legal trouble, and -above all- attract fewer Premium subs.
Upvote
2
(2
/
0)
And what drives you to believe that he will follow the constitution and laws, when his entire life has been spent flouting both? Now that he is finally in a position of great power (largely due to the religious fanatics) he routinely ignores court orders and issues insane proclamations, driving the US government directly into fascism. He wants the United States of America to become the United States of Trump, and is actively trying to destroy all who oppose him. He wants to have his likeness on new coins, Mount Rushmore, and everything in both North and South America, and eventually the entire world. Malignant narcissism combined with a 3rd or 4th grade level moron is not curable, and he has an army of very feeble-minded enablers to ensure that he is not deposed.
Upvote
0
(1
/
-1)
I'm going to push slightly against the grain. Yes, Grok shouldn't output CSAM. but users will always find a way around those safe guards. Or post content they generated elsewhere.
So why shouldn't X ban those people? I know this is a "guns don't kill people" argument. Still, those who want to dabble in CSAM need to be dealt with.
Of course I do wonder where Musk got the training data for Grok
So why shouldn't X ban those people? I know this is a "guns don't kill people" argument. Still, those who want to dabble in CSAM need to be dealt with.
Of course I do wonder where Musk got the training data for Grok
Upvote
0
(2
/
-2)
Will Grok generate images of Elon Musk in compromising positions with children? I would bet not.
Upvote
1
(1
/
0)
Well .. by extension of same logic neither does Sarin or VX nerve gas
Upvote
4
(4
/
0)
Why not run a sub pledge drive and make that a public target? If we get x number of new subscribers we'll leave X?
Also, maybe add one or more additional sub tier(s)? $25 and $50 a year are both pretty minimal compared to most other sites with subscription options. I'm sure you don't want to price subs out of reach, but why not add a $100/year option for people who can and would like to pay more? I don't know what kind of take rate The Autopian, say, gets for their higher subscription tiers, but their lowest is $50 and, as much as I like that site, Ars has a lot more and broader content that I see as more valuable to me. I'm sure you could get creative with some low-cost but unique perks like they do for higher sub tiers.
Upvote
2
(3
/
-1)
… well I shouldn’t speculate but knowing Elon’s inclinations
I bet he has that as an personal option
Last edited:
Upvote
2
(2
/
0)
Barely can do hentai stuff now on grok. I people because of it. Grok is currently super lobotizmed and almost useless anyways. All thanks to bad actors. With people like them we don't even need enemies.
Upvote
-14
(0
/
-14)
You can't do hentai on grok because the resident yandere girlfriend, Ani, is killing the image before they get to you.
Last edited:
Upvote
8
(8
/
0)
No, it's not. Photoshop does not immediately respond to a simple prompt by posting the output on twitter.
Upvote
11
(11
/
0)
Props to ft.com, who have posted an article entitled:
https://www.ft.com/content/ad94db4c...ft&token=a7473827-0799-4064-9008-bf22b3c99711
(Long link is long)
Who’s who at X, the deepfake porn site formerly known as Twitter
https://www.ft.com/content/ad94db4c...ft&token=a7473827-0799-4064-9008-bf22b3c99711
(Long link is long)
Upvote
16
(16
/
0)
Tell me exactly why you think Section 230 has any relevance here
Upvote
15
(15
/
0)
Rhurazz2012
Wise, Aged Ars Veteran
Umm, short answer, this is Elmo's fault. He knew what he was doing when he created this in the first place. Had he put in some fail safes in the first place, this would've curbed a lot of it, if not outright halted it. AI is too dangerous when put in some bad actor's hands..
Upvote
7
(7
/
0)
That's extremely messed up. There are obviously messed up people who are going to use grok that way if they can. And Elon Musk's company isn't going to even try stop them?? That's really bad
Upvote
1
(1
/
0)
Yeah, I knew this one wouldn't be received well. I've just been listening to the journalism podcast Question Everything lately and have been coming around to the idea that we need some kind of modification to Section 230 if we want to save the good of the internet from the rise of fascist slop machines.
But I'm totally open to alternative strategies that don't involve 230 changes if anyone has them. I just don't like the internet that we live in and want something better.
Suggestions welcome!
Upvote
-5
(1
/
-6)
Section 230 is irrelevant to dealing with content generated by an agent of a platform.
Upvote
18
(18
/
0)
Ah, the silliness of rules for those who don't follow rules. Wait a year. Maybe two years? There will be local models powerful enough to make explicit videos using any person, doing anything. There will be nothing law enforcement can do about this, except more knee-jerk reactionary rules and laws.
We live in a real weird time. Are you sure that's your wife calling? You're sure today. What about tomorrow? You certain that facetime chat is really them? Interesting times...
We live in a real weird time. Are you sure that's your wife calling? You're sure today. What about tomorrow? You certain that facetime chat is really them? Interesting times...
Upvote
-12
(0
/
-12)
Why in the absolute FUCK did you not follow the podcast's title Question Everything and question how you were going to get such a modification through a Republican controlled House and Senate (and even if the House is lost in the mid-terms, and the long shot that the mid-terms weren't rigged by Republicans) or not come to the conclusion by 'Questioning Everything' that forcing such reforms through was going to require winning the Presidency and then steamrolling the judicial branch to put such reforms in place using MAGA's tools and the Emil Bove Precedents?
The solution is to take down or severely backtax the AI companies not just to modify Section 230 to suddenly apply to user's AI prompts, which is what coded language suggesting 230 be amended usually entails. Neither users nor platforms should get protections on first party speech because Elon's prior modifications of Grok up to this point preclude Section 230 protections no matter what users input. Plus their TOS signed away those protections.
'Question Everything' implies knowing enough about a subject to ask an appropriate question to start with.
Upvote
-1
(0
/
-1)
Yeah, I'm a critic of Section 230, but I'm struggling to see how anyone connects it to this situation. Musk is attempting to lay the blame on the prompter for Grok generated CSAM, but that is legally indefensible.
IRL, If a person generates CSAM on spec for someone else, everyone involved has broken the law, not just the one placing the order. The second party placing the order probably is illegal (certainly the resulting possession is), but the acts of generating, possessing, and distributing the CSAM by the first party are also crimes. Section 230 may protect a site like reddit or Facebook if someone else generates and posts the content, so long as the site removes and reports the content as soon as they are aware of it. But it does not in any way protect sites that GENERATE such content, which is something Grok is actively doing.
This is not an "either, or" scenario. But a "yes, and". The prompters are likely breaking the law by asking Grok to generate CSAM, but so too is X when Grok generates the requested CSAM. And no, banning those who prompt for CSAM is not sufficient action to invoke Section 230 protections. That X cannot control the output of Grok does not absolve X of legal responsibility for what ti does. Quite the opposite, it creates a vast amount of legal liability for putting out a system they cannot control. Or at least, that is how it should be if the courts were not in the pocket of billionaires such as musk.
Upvote
8
(8
/
0)
No, the public insists that X must do age verification for all its users. Especially since its "america first voices" content is harmful to humanity.
Upvote
0
(0
/
0)
True... but how about 80%?
How many of the abusers will they successfully ban? Also not 100%.
Having an LLM available for the masses means you get responsibility. Wether you want it or not.
Musk lacks the capacities to deal with that, he probably does not realize this. In his view, everybody is just stupid. That is a much more comfortable thought.
Upvote
4
(4
/
0)
What are you talking about? There’s no general sentiment here that they shouldn’t be banned. If anything, the sentiment is people commissioning CSAM on X should face far more severe punishment than simply being banned. It’s just that xAI aren’t themselves blameless. They’re the ones generating the CSAM on request. They need to be punished and stop doing that too.I'm going to push slightly against the grain. Yes, Grok shouldn't output CSAM. but users will always find a way around those safe guards. Or post content they generated elsewhere.
So why shouldn't X ban those people? I know this is a "guns don't kill people" argument. Still, those who want to dabble in CSAM need to be dealt with.
Of course I do wonder where Musk got the training data for Grok
The “training” data is a red herring. You can legitimately obtain legitimate explicit imagery and separately legitimate, innocuous stock images of children. Using both types of data in generating the model can therefore result in it being possible to extract CSAM as an output even though none was used as an input because it can (in extremely simple terms) average out between sources within those two data types.
It’s just how this technology works. Some people see that as an excuse for the result. Others, like myself, see it as an inherent flaw conferring culpability for (but not rights to) the outputs on the model’s creators.
Upvote
8
(8
/
0)
Wrong solution to the wrong problem, which is your specialty.
I'll say it again: the output of an agentic tool of a company is not user-generated content. These tools are designed to produce the output they produce, and it is wholly valid to assert their output is simply not protected under Section 230. The user fiddling with the knobs is a fig leaf.
Upvote
6
(6
/
0)
