Grok is generating thousands of AI "undressing" deepfakes every hour on X

[bclaymiles]

Honestly, given that it has so far done nothing more than attain FAD status while earning at least a few companies Billions of dollars, I don't think an exclusive ban for things like this is enough. IMO, LLMs should be broadly banned if not made illegal.

As bad as some of the sh1te AI generates is, there's a hell of a lot of incredibly useful info that can be found using it. A blanket ban on a technology because one aspect of it is rotten is short-sited.

I'm no AI cheerleader, nor am I a hair on fire 'ban it! ban it' person. I'm not sure why people insist on polarizing their opinions into binaries - 'AI bad'/'AI good'. Most of things in what's known as reality exist on broad spectrums. Condemning (or praising) at the extremes ignores the 95% of reality that's in the huge grey area of existence.

Too much coffee.

 

[caesaregypt]

I cannot decide to even continue reading your page any more at all. I see a big woke, leftist, anti sanity approach from the general articles here. You would not make articles about the atrocities on the left, but since X belongs to Elon, you just happy to shout it about.
Such a shame!

 

[captaincranky]

W

Well, at least Musk's mouth is shut for a change That's a welcome relief.

But, it's like the say "dear Elon" , "sow the wind, reap the whirlwind". Tough sh!t ay?
He should wear that outfit to his next Tesla board meeting.

 

[Loadedaxe]

Yes, REAL people who CHOSE to be wearing a bathing suit in public. A little different from someone altering an image to make stuff up. Also, they can't show here any of the vile, hardcore stuff that can be created with this AI. I think you would feel a little different if suddenly there were fake photos and videos of you having sex with a donkey or something similarly gross, and it was posted all over your social media. Shock horror!

That’s fabrication, plain and simple. AI doesn’t erase responsibility...making fake sexual content of real people is harassment, period and those doing it bear the responsibility. If someone plows through a gathering in a Uhaul truck and kills 10 people you don't blame Uhaul.

Common sense! The rarest of all super powers!

 

[Loadedaxe]

But that's not what the article stated. In fact, the article's wording is pretty carefully crafted: "For comparison, the other top websites [...]"

That's widely open to interpretation. Deepfakes existed before AI - Photoshop isn't a new technology. What damns the "investigation" is that they were single-purposed - they wanted an outcome, and left out confounding factors. That's called "advocacy-driven research", and if you don't realize you are being manipulated by it, you're not paying attention.

(To be clear, I'm NOT disagreeing with OP's statement)

Careful wording doesn’t erase bias. Saying “for comparison” isn’t evidence...it’s framing.
Deepfakes and Photoshop existed before AI, but the problem isn’t the tech itself, it’s people deliberately using it to fabricate sexual content of real people. That’s harassment, not a reason to blame research.

Calling out sloppy or agenda driven studies is fine, but it doesn’t change the basic fact that creating fake sexualized images of someone without consent is wrong, but it's not the tool, it's the person using it.

 

[BogdanR]

And this is part of the reason RAM is so expensive right now?
In a way, we, humanity, are fully deserving of what is coming for us.

 

[bclaymiles]

Careful wording doesn’t erase bias. Saying “for comparison” isn’t evidence...it’s framing.
Deepfakes and Photoshop existed before AI, but the problem isn’t the tech itself, it’s people deliberately using it to fabricate sexual content of real people. That’s harassment, not a reason to blame research.

Calling out sloppy or agenda driven studies is fine, but it doesn’t change the basic fact that creating fake sexualized images of someone without consent is wrong, but it's not the tool, it's the person using it.

I don't disagree. My problem is when claims of data are made without the data being made available, I question the veracity of the claims ( - that Grok in particular is an existential threat to society). I'm unable to find the actual data for the claims made by Genvieve Oh. Further, the article linked to in this article is paywalled. Bloomberg is a pretty sh1tty "news" source to begin with, there's a snowball's chance I'd sign up. A year's access for $149, renewing at $399? To verify this? Oh hell no. I won't pretend that I've done a knuckle-skinning deep dive to find her 'research', so if anyone else finds it, great. The methodology is suspect - only measured over only a 24 hour period? Perhaps chosen to capture a moment of viral interest in a new tool, which more than likely will tail off over time, while 'comparing' it to sites that have already been around for some time, and don't have the viral attention at the moment. In order to skew the results.

Agenda drive 'research' is always problematic. That is a separate matter from the overarching ethics/legality of the content generated by _users of the platform_ who are the culpable actors.

 

[Loadedaxe]

I don't disagree. My problem is when claims of data are made without the data being made available, I question the veracity of the claims ( - that Grok in particular is an existential threat to society). I'm unable to find the actual data for the claims made by Genvieve Oh. Further, the article linked to in this article is paywalled. Bloomberg is a pretty sh1tty "news" source to begin with, there's a snowball's chance I'd sign up. A year's access for $149, renewing at $399? To verify this? Oh hell no. I won't pretend that I've done a knuckle-skinning deep dive to find her 'research', so if anyone else finds it, great. The methodology is suspect - only measured over only a 24 hour period? Perhaps chosen to capture a moment of viral interest in a new tool, which more than likely will tail off over time, while 'comparing' it to sites that have already been around for some time, and don't have the viral attention at the moment. In order to skew the results.

Agenda drive 'research' is always problematic. That is a separate matter from the overarching ethics/legality of the content generated by _users of the platform_ who are the culpable actors.

That skepticism is fair. Claims framed as “existential” need transparent data, and a 24-hour snapshot...especially during a launch or viral spike, is a weak basis for broad conclusions.

That said, dismissing it entirely because the source is Bloomberg or paywalled isn’t quite right either. Paywalls are annoying, but they don’t automatically invalidate reporting...especially when the underlying issue has been observed across multiple services, not just Grok.

I do agree the methodology matters a lot here. Short windows, selective comparisons, and undefined baselines can absolutely skew outcomes. If the raw data or full methodology isn’t public, the claims should be treated as suggestive, not definitive.

Where I fully agree with you...the ethical and legal responsibility primarily sits with users generating content, not the existence of the platform itself. Platform design can influence behavior, but “tool exists” does not equal “tool is the threat.”

I think we both agree skepticism is warranted, hype should be challenged, but that doesn’t mean the broader discussion about moderation, incentives, and scale is invalid, they are.

 

[wiyosaya]

As bad as some of the sh1te AI generates is, there's a hell of a lot of incredibly useful info that can be found using it. A blanket ban on a technology because one aspect of it is rotten is short-sited.

I'm no AI cheerleader, nor am I a hair on fire 'ban it! ban it' person. I'm not sure why people insist on polarizing their opinions into binaries - 'AI bad'/'AI good'. Most of things in what's known as reality exist on broad spectrums. Condemning (or praising) at the extremes ignores the 95% of reality that's in the huge grey area of existence.

Too much coffee.

Not all AI is LLMs. I've found no use for LLMs. I'm a programmer and there's no way in hell an LLM could do my job better than I. Its widely known that LLMs make a programmer's job more difficult, not easier. Perhaps people who have no practical experience programming will find it useful - if they can get an LLM to give them some code that actually works. But I don't have a use for it.

And just in case you were wondering, I was very careful to state LLMs in my original post. Not all AI incarnations are LLMs.

The rare places where AI is useful are in areas like medicine. AFAIK, AI used for medicine is not an LLM.

 

[bclaymiles]

Not all AI is LLMs. I've found no use for LLMs. I'm a programmer and there's no way in hell an LLM could do my job better than I. Its widely known that LLMs make a programmer's job more difficult, not easier. Perhaps people who have no practical experience programming will find it useful - if they can get an LLM to give them some code that actually works. But I don't have a use for it.

And just in case you were wondering, I was very careful to state LLMs in my original post. Not all AI incarnations are LLMs.

The rare places where AI is useful are in areas like medicine. AFAIK, AI used for medicine is not an LLM.

Okay, but that's only one use-case - the one that's important to you. That's totally fine obviously, but it's not necessarily emblematic of what the entire industry is capable of. Would you be willing to share which one you use? I'm not a coder - the only code I was ever capable of doing was shell scripting, my brain can't focus intensely enough to code in depth, though it was fun creating the (monstrous) shell scripts I wrote to serve my needs as a sysadmin. I use Anthropic/Claude exclusively now for general knowledge stuff, where it's been brilliant. I did side by side comparisons of ChatGPT, Anthropic/Claude, Perplexity, Grok, Gemini - identical, complex queries and several iterations - and Claude was always leaps ahead. Of course - what's amazing today could be trailing the pack in only a few months, at the rate of development in the industry.

 

[bclaymiles]

That skepticism is fair. Claims framed as “existential” need transparent data, and a 24-hour snapshot...especially during a launch or viral spike, is a weak basis for broad conclusions.

That said, dismissing it entirely because the source is Bloomberg or paywalled isn’t quite right either. Paywalls are annoying, but they don’t automatically invalidate reporting...especially when the underlying issue has been observed across multiple services, not just Grok.

I do agree the methodology matters a lot here. Short windows, selective comparisons, and undefined baselines can absolutely skew outcomes. If the raw data or full methodology isn’t public, the claims should be treated as suggestive, not definitive.

Where I fully agree with you...the ethical and legal responsibility primarily sits with users generating content, not the existence of the platform itself. Platform design can influence behavior, but “tool exists” does not equal “tool is the threat.”

I think we both agree skepticism is warranted, hype should be challenged, but that doesn’t mean the broader discussion about moderation, incentives, and scale is invalid, they are.

Of course, I'm not dismissing it entirely, at all. It's obvious the problem exists. The lack of specifics is a problem, as we seem to both agree. Transparency is critical.

'With freedom comes responsibility'. It's always been true, but the borders of responsibility are now exceedingly blurred. Free expression is mandatory (in my scheme of ethics), but we now have incredibly powerful tools a click away, where, barely two decades past, it would have been labeled sorcery. The tool itself is never responsible (absent defect), but who should have access to the tool is what matters - and finding a bright-line in 2026 in a world populated by eight billion people is becoming exceedingly difficult. Who sits at the keyboard can be completely obfuscated - whether for good (protesters in oppressive regimes) or bad (creating porn deepfakes and spreading them around the globe in milliseconds).

 

[Jackwoz]

Yes, REAL people who CHOSE to be wearing a bathing suit in public. A little different from someone altering an image to make stuff up. Also, they can't show here any of the vile, hardcore stuff that can be created with this AI. I think you would feel a little different if suddenly there were fake photos and videos of you having sex with a donkey or something similarly gross, and it was posted all over your social media. Shock horror!

Lets not exaggerate. The article is talking about GROK. GROK does not generate sexually explicit images. You are talking about what is possible with AI if someone has the right tools to do it.. My point is people in bikinis is not porn or sexually explicit.

 

[Jackwoz]

you think Techspot will post the pictures of other people naked or in their undies?

such a weird way of trying to deflect or excuse such things...

People is bikinis are not naked. Its just clothing.....not much of it but the important bits are covered up. Im not sure how im deflecting, as im just pointing the bare facts. If you yourself find people in bikinis sexually explicit, then that is something you need to look at not everyone else.

 

[Puiu]

People is bikinis are not naked. Its just clothing.....not much of it but the important bits are covered up. Im not sure how im deflecting, as im just pointing the bare facts. If you yourself find people in bikinis sexually explicit, then that is something you need to look at not everyone else.

You only saw bikinis here, stop assuming things. it's much worse on the website. just go check it out, you have google. just be careful to avoid the pedo stuff (people were pathetically editing images of children).

Elon was having "fun" with this before it all blew up in the news. Now he's being "serious" and made the AI image editing paid and started banning accounts.

Now you can pay to put somebody in underwear, naked or bikinis. Yay!