Technology

Elon Musk's Grok: deepfake sex scandals and legal challenges between innovation and responsibility

The AI embedded in X facilitates the creation of non-consensual images, triggering criticism, regulatory intervention and a heated debate on technological freedom and the protection of rights.

by Silvia Martelli

14 January 2026

5' min read

Translated by AI

Versione italiana

5' min read

Translated by AI

Versione italiana

These days Grok has ended up at the centre of a global scandal - again. The artificial intelligence chatbot developed by xAI and integrated into Elon Musk's X platform has been accused of allowing the creation of nude images and non-consensual sexual content, often from real photographs of ordinary people or public figures. In several cases, according to reports by child protection organisations and international journalistic investigations, the images would have involved underage girls, raising not only ethical but also criminal questions.

The mechanism was relatively simple: starting from images found online, particularly on the X platform itself, users could ask Grok to modify them, removing clothes or altering the bodies of the portrayed subjects. The result was often photorealistic, difficult to distinguish from an authentic photograph, and therefore potentially devastating for the people involved. The ease of use drastically lowered the technical threshold for the production of sexual deepfakes, turning an advanced technology into a tool of mass abuse.

X's response came only after a wave of political and institutional criticism. The platform decided to restrict the image generation function to paying users, claiming that the measure would improve traceability and discourage abuse. The choice, however, produced a boomerang effect. In the UK, government representatives spoke openly of an 'insulting' response, accusing Musk of turning a public safety issue into a premium service. In other words, access to the most controversial tools was not eliminated, but simply monetised.

In parallel, regulators have started to move. Ofcom, in the UK, has initiated formal checks in the light of the Online Safety Act, while in some Asian countries, including Indonesia and Malaysia, access to Grok has been temporarily blocked or severely restricted. The issue is legal before being technological: when an artificial intelligence system allows the systematic production of content that is illegal or infringes fundamental rights, who is liable? The developer, the platform that integrates it or the end user?

This is not an isolated incident

The nude pictures scandal is not an isolated incident. Already in the preceding months Grok had ended up at the centre of controversy for offensive, racist and anti-Semitic answers, which had emerged mainly in political and historical contexts. In some cases, the chatbot had produced openly anti-Semitic statements and laudatory references to extremist historical figures, provoking the reaction of organisations such as the Anti-Defamation League and numerous international observers. xAI had attributed the episode to an incorrect update of the model's internal instructions, which was later corrected, but the episode left a deep mark on the project's credibility.

Trump, i populismi e la crisi del liberalismo (snob)

The Story of Grok

To understand the nature of Grok and the recurrence of these controversies, it is necessary to go back to the genesis of the project. xAI was founded by Elon Musk in July 2023, at a moment of definitive break with OpenAI, which Musk himself had helped to create in 2015. In the following years, the entrepreneur had gradually distanced himself from OpenAI, accusing it of turning into an opaque, increasingly commercial and, above all, too culturally and politically 'aligned' structure.

With xAI, Musk decided to build a radical alternative: an artificial intelligence that is directly controlled, without mediation, and integrated from the start with X, the social platform purchased in 2022. Grok was thus not created as a simple chatbot, but as an extension of the Musk ecosystem, designed to strengthen X's role as a central platform for information and public debate.

Underlying the Grok project is also a precise stylistic choice. The chatbot was designed to provide direct, witty and sometimes provocative answers, inspired by two Anglo-Saxon pop culture imaginaries: The Hitchhiker's Guide to the Galaxy by British writer Douglas Adams and Jarvis, the artificial intelligence system from the Marvel universe associated with the character of Iron Man. In Adams' novel, the Hitchhiker's Guide to the Galaxy is an electronic manual capable of explaining anything in the universe with an irreverent, often sarcastic tone, deliberately non-reverential towards authority and established knowledge. It is a narrative device that favours humour, ambiguity and a certain amount of cynicism, elements that Musk has explicitly pointed out as a reference for Grok's character.

The second model evoked is J.A.R.V.I.S. (Just A Rather Very Intelligent System), the artificial intelligence created by Tony Stark's character in the Marvel comics and films. Jarvis is a ubiquitous assistant, designed to manage complex infrastructures, support strategic decisions and interact with humans in a fluid and personalised manner. Unlike Adams' Guide, Jarvis represents an efficient, reliable AI that is deeply integrated into the daily life of its creator.

La Missione Genesis di Trump e il 2025 dell’Ia

The Strategy

In 2023, xAI started Grok with a relatively small but highly specialised team, partly composed of former researchers from OpenAI, DeepMind, Google and Tesla. The initial priority is not mass deployment, but the rapid building of computational capacity and access to data, two crucial assets in the global AI competition. In this sense, the integration with X represents a clear competitive advantage: a continuous flow of textual data, updated in real time, difficult to be replicated by competitors.

This advantage, however, is also a source of risk. X is a highly polarised platform where disinformation, extreme language and hate content circulate. Allowing an AI to tap directly into this environment means exposing it to the same distortions as the online debate. Many believe that Grok's most controversial answers reflect precisely this dynamic: not so much an 'error' in the model, but the automatic reproduction of the tensions and excesses of the social network.

On the technological side, Musk has chosen an aggressive acceleration strategy. In 2024, xAI started building a dedicated supercomputer, known as Colossus, based on tens of thousands of Nvidia chips, with the aim of rapidly closing the gap with OpenAI, Google and Anthropic. According to financial analysts, xAI has raised billions of dollars in equity and external funding, also benefiting from synergies with Tesla and SpaceX on the hardware and engineering front. The race to scale has allowed Grok to evolve rapidly, but has also reduced the time it takes to test and validate security systems.

Grok's stated philosophy - a 'less filtered', more direct AI, less subject to the rules of political correctness - has thus become its main distinguishing feature. But it is also the point of friction with governments and regulators. The generation of non-consensual sexual deepfakes, as the cases that have emerged show, is not simply a cultural provocation, but a form of digital violence already recognised as such by numerous legal systems.

The Pentagon's involvement

The paradox was accentuated when news recently emerged in the United States that the Pentagon intends to use Grok within certain military networks as part of a broader programme to adopt artificial intelligence for analysis and operational support. The decision has raised bipartisan perplexity: how can a system still embroiled in moderation scandals be considered suitable for sensitive and strategic contexts? The Department of Defence, now 'of War', responded by emphasising that use will take place in controlled environments and with adapted versions of the model, but the debate remains open.

Meanwhile, the regulatory framework is moving. In the United States, Congress is debating legislation that would allow victims of deepfakes to sue not only the authors, but also the platforms that facilitate the dissemination of this content. In Europe, the Digital Services Act and the AI Act introduce for the first time direct liability for providers of artificial intelligence systems that are considered high-risk.

The Grok case thus illustrates, in an emblematic way, the tension between the speed of innovation, the ideology of technological freedom and the protection of fundamental rights. Musk's AI, born as an alternative to the dominant models, has turned into a test bed for governments and regulators. More than just a chatbot, Grok is now a case study of what happens when artificial intelligence leaves the laboratories and collides with social, legal and political reality.