Elon Musk’s AI company xAI has introduced measures to tighten controls on its Grok image generation feature. The decision comes after growing concern around how users on the X platform were editing photos of real people.
Regulators, rights groups, and the media have raised alarm about the misuse of the tool to generate non-consensual deepfake pornography. The backlash has been unanimous from America to Europe and across Africa.
Reports from authorities and journalists showing Grok being used to generate explicit content without consent prompted xAI to act.
According to the company’s safety team, new safeguards now block all users from editing images of real people into revealing clothing such as bikinis or underwear in jurisdictions where such content breaches local law. The move applies to everyone, including premium subscribers.
The change marks a significant shift for Grok. Earlier complaints had described how users could prompt the AI to “undress” or alter photos to create deepfake imagery of women and minors. Investigations in California say more than half of tens of thousands of images generated over a recent holiday period showed people in minimal clothing, with some appearing to be underage. That legal scrutiny added weight to the criticism.
xAI’s Grok Image Features Now Under Tighter Control
In a post on its official safety handle, X stated that while Grok’s image features are now tightly controlled, they have not been entirely removed. Image creation and editing through Grok will still be available, but only to paid subscribers on the X platform.
The company said this would improve accountability and help trace misuse. Geoblocking has also been introduced in countries where such creations are illegal, effectively stopping users in those regions from generating problematic content.
The tightened controls have not silenced critics. Some regulators and watchdogs have said the measures still fall short.
Media reports and investigative testing suggest that safeguards can be bypassed in some cases. Critics argue that paywalling harmful capabilities does not prevent their existence but rather stacks them behind a subscription wall. Others note that the standalone Grok app and web portal still allow certain image manipulation without the same restrictions.
There is also widening international pressure. Government officials in California, the European Union, and several Asian countries have all voiced concern about deepfake misuse.
Some regulators are now considering or pursuing legal action against xAI and X for not doing enough to prevent predatory or exploitative imagery from being created on their platform. In response, the European Commission has said it will assess the new safeguards to ensure they truly protect citizens.
This episode comes amid a broader debate about how to govern generative AI responsibly. An earlier analysis highlighted that Grok’s “unfiltered” design made it particularly vulnerable to abuse, turning it into a tool for non-consensual deepfake pornography at scale. That article explored how the combination of lax safeguards and powerful generative models can rapidly become a social and legal crisis if unaddressed.
For now, xAI maintains it is adjusting Grok in response to emerging risks. The company has framed the latest controls as part of ongoing efforts to work with users, regulators, and partners on safety.
However, the broader challenge remains: how to balance innovation in generative AI with robust protections for individuals’ privacy, dignity, and rights. Regulators and advocacy groups say this is only the beginning of a complex legal and ethical journey.