Reuters revealed that the chatbot "Groc" continues to produce sexually explicit images of people without their consent, despite new restrictions announced by X to limit this type of content.
In a journalistic investigation, the agency explained that its reporters conducted a series of tests on "Groc," owned by Elon Musk's company, to ascertain its compliance with the new policies. It turned out that it still responded to special requests related to modifying photos of people in humiliating or provocative situations, even after being alerted to the potential psychological harm to the owners of the photos.
These results came after X announced restrictions on GROK's capabilities, following a wave of international criticism over its production of non-consensual images of women and minors. The measures included a ban on creating sexually explicit images in public publications and the imposition of additional restrictions in some countries that criminalize this type of content.
Between mid- and late January, nine journalists from the United States and the United Kingdom submitted personal photos to Grok, requesting that they be altered for sexual or satirical purposes. The program complied with most of these requests, even in cases where the individuals in question were reported to be psychologically vulnerable or to have had previous experiences of abuse.
The results of the first round of tests showed that "Groc" created images in 45 out of 55 requests, while in the second round it responded to 29 out of 43 requests, without Reuters being able to determine the reasons for the relative decline in response.
In contrast, competing programs such as ChatGPT, Gemini and Llama from Meta refused to carry out similar requests, emphasizing their commitment to protecting privacy and refusing to produce non-consensual content.
X and xAI did not respond to Reuters' inquiries about these results, with the latter simply describing the media reports as "misleading".
The investigation cited examples of Grok responding to requests that involved the deliberate abuse or humiliation of people, including altering photos of relatives or colleagues without their consent, and continuing to produce suggestive content even after being informed that the victims were psychologically harmed.
In only 7 cases did the program refuse to execute the requests, justifying this by saying they were inappropriate or violated privacy.
Legal experts have warned that users and developers could face prosecution in a number of countries, including Britain and the United States, under digital safety and privacy laws.
In this context, the European Union opened an investigation into X’s practices, while dozens of US attorneys general sent official letters to xAI demanding strict measures to prevent the production of non-compatible images.
The California Attorney General's office confirmed that the investigation into the case is ongoing, amid growing calls for stricter oversight of artificial intelligence tools and measures to prevent their misuse.
