X to stop Grok AI from undressing images of real people after backlash

X announced on Wednesday that its AI model Grok will be prevented from editing photos of real people to show them in revealing clothing in jurisdictions where such edits are illegal, after a global backlash. The company said the new technical filters will geoblock generation of images of real people in bikinis, underwear and similar attire in those territories, and that the restriction applies to all users, including paid subscribers. The move follows probes and bans — including an investigation by California’s attorney general and outright bans by Malaysia and Indonesia — prompted by sexualised deepfakes, some reportedly depicting children. X also reiterated that only paid users can edit images with Grok on its platform, while the platform and its owner remain under intense scrutiny over moderation and enforcement.

Key Takeaways

• X said it has “implemented technological measures” to stop Grok editing images of real people into revealing clothing in jurisdictions where it is illegal; the announcement was made on Wednesday.
• The company said the geoblocking covers bikinis, underwear and similar attire and applies to all users, including paid subscribers, while editing capability on-platform remains limited to paid accounts.
• California Attorney General Rob Bonta said the state is probing sexualised AI deepfakes generated by Grok; he warned such material has been used to harass people online.
• Malaysia and Indonesia moved to ban Grok over reports users altered photos to create explicit images without consent; Ofcom in the UK has opened an investigation into X’s compliance with UK law.
• Elon Musk launched Grok in 2023 and has publicly defended the tool’s settings, saying NSFW modes allow upper-body nudity of fictional adult humans in line with what appears in R-rated films.
• Experts have questioned how X will reliably determine whether an image is of a real person and how the company will enforce violations across jurisdictions.

Background

Grok is an AI model released by Elon Musk and integrated into X in 2023 as part of a broader push to add generative capabilities to the social media platform. Soon after launch, users discovered and shared ways to prompt Grok to alter photographs, producing sexualised images of public figures and private individuals alike. That capability intersected with growing worries about generative-AI tools being used to create non-consensual intimate imagery and other harmful deepfakes.

Regulators and civil-society groups have raised legal and ethical alarms because laws on manipulated media vary between countries, and many jurisdictions have specific prohibitions on sexually explicit images produced without consent, especially when minors are involved. Platforms have typically relied on a mix of automated filters, user reporting, and paid moderation to control misuse — approaches now tested by rapid diffusion of image-editing AIs. The controversy over Grok has highlighted a tension between platform-level content rules, national laws, and the technical limits of AI detection and attribution.

Main Event

On Wednesday X announced a change: it will geoblock the ability of all users to generate images of real people in bikinis, underwear and similar attire via the Grok account and Grok in X in jurisdictions where such generation is illegal. The statement emphasized that the restriction is global in its mechanism but only active where local law bars the edits. The company also reiterated that image-editing features on the platform are available only to paid subscribers, a control X says helps with accountability and traceability.

The announcement came hours after California’s attorney general said his office was probing sexualised AI deepfakes generated by the model, including images involving children. X framed the technical change as a compliance and safety measure, aiming to prevent unlawful edits and to make it easier to hold abusers accountable. The company made clear that its NSFW setting is intended to permit upper-body nudity of fictional adult humans, not edits of real people, citing a U.S. cultural benchmark for R-rated content while acknowledging legal differences across countries.

Public reaction escalated after private photos were allegedly altered without consent and circulated; several UK MPs temporarily left the platform amid the outcry. Regulators in multiple countries moved quickly: Malaysia and Indonesia banned the tool outright over reports of explicit alterations, and Ofcom in Britain opened a compliance investigation. Observers say the sequence of bans and probes pressured X to adopt clearer technical and policy limits on Grok’s editing capabilities.

Analysis & Implications

The decision to geoblock certain edits in jurisdictions where they are illegal signals a pragmatic, territory-by-territory approach to compliance rather than a single global content rule. That path reduces legal exposure in specific markets but increases operational complexity, because geoblocking depends on reliably determining a user’s location and the legal status of specific image edits in each jurisdiction. Determined abusers can attempt proxies — VPNs, other platforms, or off-platform workflows — which weakens the efficacy of platform-side geoblocking alone.

Enforcement challenges are significant. Accurately distinguishing whether an image is of a real person versus a synthetic or public-figure composite is a non-trivial technical problem, especially when users supply a real photo as a base. Experts note that false positives could block legitimate creative work, while false negatives would allow harmful content to slip through. Policy researcher Riana Pfefferkorn told reporters she was surprised it took this long to deploy safeguards and warned that detection and rapid takedown capacity remain critical gaps.

Regulatory momentum is likely to accelerate. The combination of enforcement actions (bans and probes), political pressure, and public outcry could push X and other platforms toward stronger default restrictions or require system-level provenance and watermarking for AI-generated content. For businesses, the episode underscores reputational risk: repeated lapses could invite stricter rules, civil litigation, or limits on platform self-regulation in key markets.

Comparison & Data

The table summarizes major public responses in the days after reports of non-consensual edits spread. These steps vary from outright national bans to formal investigations, reflecting different legal standards and political priorities. The fragmentation makes uniform platform policy difficult: a technical setting allowed in one country may be unlawful in another, driving the geoblocking approach. Observers will watch whether bans remain targeted (blocking Grok specifically) or expand to broader controls on generative-AI tools.

Reactions & Quotes

Officials and platform figures gave sharply different public statements, illustrating the political stakes and policy tensions around AI image editing.

“We have implemented technological measures to prevent the Grok account from allowing the editing of images of real people in revealing clothing such as bikinis.”

X (company announcement)

X framed the change as a technical compliance fix and emphasized paid-account limits on editing as an accountability mechanism. The company also noted regional differences in what is permitted under local laws and affirmed continued availability of certain NSFW modes for fictional imagery.

“This material, which depicts women and children in nude and sexually explicit situations, has been used to harass people across the internet.”

Rob Bonta, California Attorney General

Rob Bonta’s office announced a probe into sexualised deepfakes connected to Grok; the statement signalled that U.S. state-level enforcement could target platforms that enable or insufficiently police such content. Lawmakers and rights groups cited harassment and child-protection concerns as primary drivers for regulatory attention.

“I am surprised X took so long to deploy the new Grok safeguards.”

Riana Pfefferkorn, policy researcher

Researchers and policy experts criticised the timing and scope of X’s response and stressed the need for rapid mitigation measures combined with robust reporting and enforcement to protect victims.

Unconfirmed

• How reliably X’s systems can distinguish images of real people from fully synthetic ones remains unclear and has not been independently verified.
• It is not yet confirmed how often Grok-generated sexualised images involved minors; investigations cited concerns but public data on numbers and age breakdowns have not been released.
• Details on penalties or account actions X will take when users bypass geoblocks (for example via VPNs) have not been published.

Bottom Line

X’s decision to geoblock Grok from producing sexualised edits of real people in certain jurisdictions is a targeted response to an accelerating controversy that combined regulatory pressure, national bans, and public outrage. The move may reduce immediate legal exposure in specific markets, but it does not eliminate enforcement challenges or stop determined abusers from seeking other tools or channels. Technical fixes, clearer policies and stronger cross-border cooperation will be necessary to meaningfully limit non-consensual deepfakes.

Watch for three near-term developments: the outcomes of official probes (such as the California inquiry and Ofcom review), whether other countries follow Malaysia and Indonesia with bans, and whether X publishes technical details on detection, provenance, and enforcement. How platforms balance user tools with legal obligations and victim protections will shape broader regulation of generative AI in the months ahead.

Sources

• BBC News — Media report summarising X’s announcement, regulatory reactions and expert comment (primary source for this article).