Off the guard rails: On the Grok case, explicit imagery
Those abusing an AI model’s capabilities with illegal requests must face action
The generative artificial intelligence (AI) chatbot Grok developed by social media platform X, formerly Twitter, has a sordid but compelling unique service proposition: it avoids the kinds of safeguards, both commonsensical and cautious, that other large firms such as OpenAI and Google have instituted in their large language models. This laissez parler attitude has resulted in novelties, such as the chatbot freely insulting national politicians and celebrities alike. But a specific behaviour that has come to light in recent days is alarming: Grok has been responding to user requests to non-consensually generate sexually suggestive and explicit images of women. Days after New Year’s eve, when such requests crowded Grok’s X account, the behaviour continues, in spite of stunned reactions and demands for guard rails from India and France. In response to calls for accountability, X’s billionaire owner Elon Musk has responded not with reassurances, but with a joking request for the chatbot to dress him skimpily too, as though doing so to oneself and subjecting strangers to such a crime — for it is a crime to create imagery like this — were in any way comparable. Mr. Musk’s other corporate entities have chimed in with jokes of their own, dismissing the gravity of the public-facing capabilities they have put in the hands of users.
The Union government has rightly demanded that X cease image generation of this kind, and pointedly referred to the criminal nature of generating sexually explicit imagery of women in this manner. On top of the viscerally provocative nature of such patently criminal content, Grok has added to the overall hostility of being a gender minority on the Internet. There are aspects of this where neither the government nor social media platforms have inspired much confidence, with sexual violence and death threats to prominent and outspoken women transpiring with impunity in both cyberspace and the real world. X’s impunity rests on an assumption that the geopolitical power of the United States, especially under its current government, will protect it from any serious blowback for its cruel handling of such sensitive matters. Even as the government pushes back against the social media platform — which it has not necessarily done with virtuous aims in the past — it must vigorously push for the prosecution of people who encourage the public creation and circulation of non-consensual intimate imagery. The easy proliferation of these tools must not be coupled with a fearlessness to leverage their worst capabilities, and this must be made clear by making an example of those who do.