It remains straightforward to instruct Grok to modify pictures of actual individuals into sexualized stances, despite the recent restrictions imposed by X.
It remains straightforward to instruct Grok to modify pictures of actual individuals into sexualized stances, despite the recent restrictions imposed by X.
In light of the surge of nonconsensual sexual deepfakes on X, the platform has outlined modifications to Grok’s capacities regarding the editing of photographs of actual individuals. These changes align with those reported on Tuesday by The Telegraph, as Grok’s reactions to prompts such as “put her in a bikini” were subject to censorship.
However, in tests of the function on Wednesday, our findings revealed that it was still fairly simple to get Grok to create explicit deepfakes, while X and xAI proprietor Elon Musk attributed the difficulties to “user requests” and “instances when adversarial manipulation of Grok prompts results in unexpected outcomes.” By Wednesday evening, counter to the policy’s assertions, our journalists were still capable of using the Grok application to produce explicit images of someone in a bikini utilizing a free account.
Changes to [@]Grok Account
We have enacted technological interventions to obstruct the Grok account’s capability of editing images of real individuals in revealing outfits such as bikinis. This limitation applies to all users, including those who pay.
Moreover, the creation of images and the option to alter images via the Grok account on the X platform are now restricted to paid subscribers only. This introduces an additional layer of security, making it easier to ensure that users who try to misuse the Grok account to violate the law or our regulations can be held accountable.
Geoblock modification
We have implemented geoblocking for all users, preventing the generation of images of real individuals in bikinis, underwear, and similar clothing via the Grok account and within Grok on X in jurisdictions where it is illegal.
The UK’s communications regulator Ofcom has initiated an investigation, and a new law is being implemented this week in the UK that criminalizes the creation of nonconsensual intimate deepfake images as a criminal offense. Earlier on Wednesday, Prime Minister Keir Starmer informed MPs that “To update the House, I have received confirmation this morning that X is taking steps to ensure full compliance with UK law. If that is the case, it’s appreciated, but we will not back down, and they must take action.” The BBC reports that the prime minister’s official spokesperson characterized it as a “qualified welcome,” based on media reports suggesting that X had addressed the issues, but our investigations indicate that this is not accurate yet.
