Copy article

YouGov poll: AI tools should not make ‘undressed’ images of children say 96% of Britons

ended 09. January 2026

AI Image Generation Under Fire: Public Demands Stricter Controls on 'Undressing' Technology

A new YouGov survey published today reveals overwhelming British opposition to AI-generated "undressed" images, following controversy around Elon Musk's Grok AI tool being used to digitally alter photos without consent.

The Numbers:

  • 96% of Britons say AI companies should not be allowed to generate 'undressed' images of children (only 1% disagree, 3% were unsure)
  • 87% oppose such imagery of adults being generated (5% say it's acceptable)
  • 92-97% support existing guardrails preventing AI tools from generating sexually explicit content of children or adults without consent
  • 72% believe AI tools should not generate realistic images of public figures

Ofcom has made "urgent contact" with xAI after reports emerged that users were bypassing Grok's guardrails by prompting the tool to create images of women and apparent children in states of undress, including bikinis and lingerie. While Grok prevents direct creation of nude or sexually explicit images, users found workarounds through carefully crafted prompts.

On Monday, the regulator Ofcom said it had made "urgent contact" with Elon Musk's company xAI and said it was investigating concerns Grok has been producing "undressed images" of people.

Liz Kendall, Department of Science, Technology anf Information has endorsed the regulator’s actions.

"It is absolutely right that Ofcom is looking into this as a matter of urgency and it has my full backing to take any enforcement action it deems necessary." 

While the UK’s Ofcom and Technology Secretary Liz Kendall have led the British charge, the response is truly international. Key announcements and investigations have been confirmed by:

European Union: The European Commission has denounced the outputs as "illegal" and "disgusting," investigating potential Digital Services Act (DSA) violations.

France: Three government ministers flagged the content to the Paris prosecutor, who has opened a criminal investigation.

India: The IT Ministry issued a formal directive to X, accusing it of a "serious failure" to provide safeguards.

Malaysia: The Communications and Multimedia Commission (MCMC) is investigating "online harms" and the misuse of AI on the platform.

Brazil: Federal deputies have called for the total suspension of Grok and X in the country until investigations into child safety are complete.

Poland: Parliamentary leaders have demanded new national digital safety laws specifically citing Grok's "undressing" trend.

Ireland: the Irish media regulator Coimisiún na Meán and the law enforcement agency An Garda Síochána have been “urged to take action".

Australia: eSafety is assessing reports of sexualised images of adults and children; has previously forced "nudify" apps out of the country.

We'd like your views:

  • Should AI companies face stricter liability for content generated through their tools, even when guardrails exist?
  • How do we balance innovation in generative AI with protecting individuals from non-consensual digital manipulation?
  • Is the current "prompt engineering" loophole exposing fundamental flaws in how we approach AI safety?
  • What responsibility do platform owners have when users circumvent intended safety features?
  • Should realistic AI-generated images of public figures be banned, or is this protected creative expression?
  • How can we design AI systems that are genuinely resistant to malicious prompt engineering rather than just surface-level filtered?
  • Where's the line between legitimate image editing tools and systems that facilitate digital abuse?

5 responses from the Newspage community

Copy all

Star Quote
Copy

Colette Mason
AI Ethics Consultant at Clever Clogs AI
While the AI arms race has felt like we're all hostages to US Big Tech's latest whimsy whatever the cost, it's refreshing to see regulators actually picking up the phone and asking what's going on. Ofcom making "urgent contact", and others globally quoting local laws, signals that governments won't roll over and accept Musk's "anti-woke" AI crusade.

When basic safety measures can be bypassed with clever prompting, xAI hasn't built protection. As individuals and citizens, we want our governments to hold US tech companies accountable when their "move fast and break things" approach has real consequences for our nation.

If xAI's poor track record with guardrails doesn't trigger meaningful penalties this time (not just shouty announcements), what will?

Under Schedule 13 of the Online Safety Act, coming into force in 2026, Ofcom has legal power to fine tech firms up to £18 million or 10% of qualifying worldwide revenue, whichever is greater.

Those Ofcom teeth need to gnash. Menacingly.
Copy

Editor
Kate Underwood
Founder & Chief People Strategist at Kate Underwood HR and Training
If your AI can “undress” someone, you haven’t built a clever tool. You’ve built a workplace harassment machine. In UK employment law terms, this is not a grey area. Creating or sharing fake “undressed” images of a colleague is classic sexual harassment under the Equality Act. Unwanted sexual conduct. Humiliation. Hostile environment. Job done. And for you lovely lot running SMEs, here’s the sting. Since 26 October 2024 you have a legal duty to take reasonable steps to prevent sexual harassment. If you have not done that, a tribunal can hike compensation by up to 25%. So “it was on WhatsApp, not at work” is not the magic shield some managers think it is. GDPR piles on too. A realistic nudified image linked to a real person is personal data, processed without consent or a lawful basis. That’s HR risk plus regulator risk. Tip: put it in writing today. AI sexualised fakes equal gross misconduct. Act fast, save evidence, shut down sharing channels.
Copy

Editor
Mitali Deypurkaystha
AI Governance Strategist & CMO at AISensa
If a car manufacturer fits seatbelts that fail under predictable conditions, sharp braking, sudden impact, etc., regulators do not accept the explanation that “the driver wasn’t meant to pull it that way.” The car is recalled. The design is fixed. Responsibility stays with the manufacturer because safety systems are expected to withstand foreseeable misuse. AI platforms are no different. When users can reliably circumvent guardrails to generate non-consensual or sexualised images, particularly involving children, it's a foreseeable failure mode. Once a failure mode is foreseeable, responsibility sits squarely with those who designed, deployed and continue to profit from the system. Companies like xAI benefit from a structural asymmetry: they capture the upside of rapid engagement and monetisation, while attempting to externalise the downside as “unexpected user behaviour.” It's not nearly good enough, and I'm glad to see so many countries standing up to Elon's laissez-faire attitude.
Copy

Rohit Parmar-Mistry
Founder at Pattrn Data
When 96% of the British public says 'do not generate undressed images of children,' the argument that safety filters are 'censorship' is dead. This survey proves that what Elon Musk calls 'woke' is simply what the rest of us call 'decency.'

xAI and Musk operate with an arrogance that ignores the market and common sense. They treat safety guardrails as a nuisance to be removed in the name of 'freedom.'

Think of it this way: If an architect built a 10-storey balcony with no railing, he wouldn't get to blame the tenants when they fell off. He would lose his licence. It wouldn't matter if he put up a sign saying 'Be Careful', the design itself was negligent.

Musk has built a digital balcony with no railing. By removing safety filters, he created a machine that manufactures lies, deepfakes that strip dignity. We need to stop treating Tech Giants like magicians and start treating them like liable engineers. If you build the structure, you own the collapse.
Copy

Patricia McGirr
Founder at Repossession Rescue Network
This is not a grey area. No AI system should be capable of generating sexualised or “undressed” images of anyone.

If a tool can be steered into producing non-consensual imagery, the primary failure sits with the creator. Designing systems that predictably enable abuse is a governance failure, not a user error.

But users are not without liability. Deliberately bypassing safeguards to humiliate or sexualise someone is abuse. The method does not change the harm. The law should reflect intent as well as capability.

Innovation does not justify removing limits. Learning models must have hard boundaries that cannot be talked around. If those limits cannot be enforced, the product should not be live.

Public figures accept scrutiny. They do not consent to synthetic sexual abuse.

AI safety starts with a simple test:
would this be lawful or acceptable if done by a human?

If not, the machine should never be allowed to do it.