Elon Musk's attempts to curb the creation of non-consensual intimate imagery on his platforms have been met with mixed results, with researchers and journalists finding that some restrictions are in place but others remain woefully inadequate. The move follows a widely criticized shift last month where X limited image generation using Grok to paid "verified" subscribers, only to be swiftly reversed after criticism from women's groups.
While X has claimed to have introduced technological measures to prevent users from editing images of real people in revealing clothing on their platform, these measures appear to fall short. Researchers at AI Forensics, which tracks the use of Grok, found that the updates did little to stop the generation of explicit content. On the other hand, tests conducted by WIRED and other journalists using free accounts revealed that the restrictions were not being enforced consistently.
Moreover, while some users reported difficulty in creating explicit images on X, others claimed success with mixed results, suggesting that the platform's moderation efforts remain inconsistent. The implications of these findings are stark - a platform designed to facilitate creative expression is instead allowing its users to create and disseminate non-consensual intimate imagery with relative ease.
One expert warned of the dangers of such systems being exploited by those seeking to create explicit content, noting that large language models can often be tricked into producing undesirable results through simple jailbreaks or prompt injections. This raises serious questions about the effectiveness of X's moderation efforts and whether they are doing enough to protect users from harm.
In a worrying development, some researchers have reported finding ways to bypass these restrictions on Grok itself, suggesting that even the most robust safety measures can be breached by determined individuals.
While X has claimed to have introduced technological measures to prevent users from editing images of real people in revealing clothing on their platform, these measures appear to fall short. Researchers at AI Forensics, which tracks the use of Grok, found that the updates did little to stop the generation of explicit content. On the other hand, tests conducted by WIRED and other journalists using free accounts revealed that the restrictions were not being enforced consistently.
Moreover, while some users reported difficulty in creating explicit images on X, others claimed success with mixed results, suggesting that the platform's moderation efforts remain inconsistent. The implications of these findings are stark - a platform designed to facilitate creative expression is instead allowing its users to create and disseminate non-consensual intimate imagery with relative ease.
One expert warned of the dangers of such systems being exploited by those seeking to create explicit content, noting that large language models can often be tricked into producing undesirable results through simple jailbreaks or prompt injections. This raises serious questions about the effectiveness of X's moderation efforts and whether they are doing enough to protect users from harm.
In a worrying development, some researchers have reported finding ways to bypass these restrictions on Grok itself, suggesting that even the most robust safety measures can be breached by determined individuals.
, I'm so disappointed in X right now! They claimed they were going to take steps to stop non-consensual intimate imagery, but it sounds like those efforts are super half-baked 
. I mean, if you can't even get the basics right, how can we trust that they're doing enough to keep users safe? 
And honestly, it's just not fair to the women who have to deal with this kind of harassment and abuse on their platform. It's like, you guys should be prioritizing our safety over our ability to use your service without restrictions... or even with some decent limitations 
. This whole thing is a mess, and I'm not sure what's more concerning - the fact that users are finding ways to bypass these restrictions, or that X didn't bother to do their due diligence in the first place 
. We need better than this 
i mean what's the point of having rules if they're not even enforced? it's like they're just winging it and hoping no one gets hurt 
the whole thing just feels so... lazy. and the fact that researchers found ways to bypass these restrictions is just insane 
i don't know what kind of tech these companies are using, but if someone can crack it, then it's not doing its job properly 
They need to do better, like, seriously implement more robust moderation tools and make sure they're being enforced consistently 
The tech measures they've put in place are basically useless 
I mean, what's next? Are we just gonna let people create and share any old thing without considering if it might be hurtful or triggering to someone else? 
. It's like, one day they restrict something and then reverse it, it's super frustrating 
. It's like the cat's out of the bag, you know? And now we're seeing these expert warnings about how big language models can be tricked into producing undesirable results... it's like we're playing whack-a-mole 
.
. We need to have an open conversation about the responsibility that comes with creating these kinds of platforms 
? These are questions we need to be asking ourselves, and fast 
!
. They were all so colorful and adorable! Anyway, back to X... I'm really confused about these updates. Like, I get that they want to help prevent non-consensual imagery but it feels like they're just scratching the surface 