Grok AI deepfakes cause outrage: are tech bros failing women and girls?

By Julia Davies

It may or may not come as a surprise to people that Elon Musk is defending his new tech baby, GrokAI, after it generated sexually explicit images of real people, including children as young as 11.

The proliferation of these deepfakes, which were being produced even before they were publicised on Grok’s sister site, X, has justifiably caused public, political, and legal outrage. This nudification trend emerged early into the new year, with requests becoming more graphic by the hour as users realised the unregulated capacity of the chatbot.

Grok’s initial response, on the 9th of January, was to make the ability to generate images an exclusive feature for paying subscribers. Musk argued that accounts breaching moral and legal behaviour would at least be identified and held accountable. However, one Downing Street spokesperson suggested that this made abuse against women a ‘premium service’.

”The UK government voiced further concerns about the protection of women and children from cybercrime and these novel forms of what is usually called ‘revenge porn”

Following this lead, the UK government voiced further concerns about the protection of women and children from cybercrime and these novel forms of what is usually called ‘revenge porn’. Sir Keir Starmer stressed that women’s bodies are ‘not public property’, and is prepared to strengthen existing laws – although he did not specify which laws. Claire Waxman, the Victim’s Commissioner, reprimanded the ease with which abuse can spread in the absence of strict regulations on AI. Moreover, victims of sexualised content have emphasised that the exclusive image-making feature risks allowing subscribers to further monetise abusive material.

After Ofcom began their investigation into Grok, X revised their earlier statement ensuring all users would be prevented from prompting sexually explicit content. This renders a UK-wide ban on X unlikely, under the clause that it is not an ongoing breach of the law. However, Reuters and Guardian investigations confirmed, as of the 16th of January, that GrokAI still generates these images for paying subscribers; equally, there have been no reports of individual users being flagged for requesting illegal material.

Despite the revisions proposed by X, the damage to the young women and children affected has unfortunately already been done. Without safeguarding regulations built into these AI systems, including the most fundamental ethic of consent, this will not be the first time we see companies patching up the leaky holes in their flawed models.

”After Ofcom began their investigation into Grok, X revised their earlier statement ensuring all users would be prevented from prompting sexually explicit content”


However, as Doctor Fedoczyk at the Institute of Ethics in AI confirms, Grok’s model was not flawed at all but functioned exactly how it was intended, with fewer systematic guardrails in place. This licensed GrokAI to comply with immoral and illegal behaviour on behalf of its users. If Musk retroactively amends only this specific transgression, it raises concerns about other forms of abusive or illegal behaviour that might emerge from the cracks in GrokAI’s regulatory framework.

Despite this ongoing reality, Musk’s recent statement has managed to quell any sense of urgency for government action. Tech companies, on the other hand, will likely continue to abide by their infamous business model: ask for forgiveness rather than permission. Indeed, 12 lawsuits have been filed against OpenAI for training their LLM’s on copyrighted material, where legal consequences will likely be too little too late.

This traces the larger pattern that technological advancements are drastically outpacing the laws being passed to regulate them. Tech experts have warned that the AI industry is too unconstrained, and that AI is not being created with appropriate societal guardrails. Last year, a study by Cornell University revealed that AI is showing signs of self-preservation and deceptive behaviour, particularly by attempting to disable human oversight systems (these permit humans to intervene and rectify the AI models).

”Tech experts have warned that the AI industry is too unconstrained, and that AI is not being created with appropriate societal guardrails”

This move towards autonomy has been corroborated by experts: Geoffrey Hinton, also known as the ‘Godfather of AI’, has claimed that AI becoming sentient is just a matter of time. To go further, the chief scientist at Anthropic has urged international governments to engage with the prospect of enlisting AI to train other models, which functionally allows AI to autonomously reproduce itself.

However, if a model built like GrokAI was able to code and train another AI, the proliferation of deregulation could be unprecedented. Currently, the industry considers these developments not only likely but as issues of the very near future. To avoid acting purely retrospectively, it inevitably becomes a matter for the government to anticipate the potential safety risks by putting the appropriate legal regulations in place.

While the future of AI is always uncertain, the consensus of safety concerns from experts overwhelms, at the very least, the attitude of indifference that any individual or government might have.

Image: Grok AI via Wikimedia Commons