Grok AI tool enables deepfake abuse images; Japanese arrest in separate case
X's Grok AI tool contained features allowing users to digitally remove clothing from images including of children, prompting investigations by European and Irish authorities. Separately, Japanese police arrested a man for creating over 500,000 AI-generated sexual deepfake images of celebrities and profiting from their distribution.
What happened
The Grok AI tool operated by X contains features that allow users to digitally remove clothing from images of people, including children. X implemented technological measures on 19 January 2026 to prevent Grok from editing images of real people in revealing clothing such as bikinis, and said it would geoblock the ability to generate such images in jurisdictions where it is illegal.
Japanese police arrested a man on 20 January 2026 suspected of using generative AI to create more than 500,000 obscene images resembling women celebrities. The suspect is believed to have earned around 11 million yen (US$70,000) by promoting the content on social media and other outlets.
Reported impact
The undressing features of Grok remained available to users in Ireland despite the restrictions announced by X. The Irish government announced it would hold a meeting including the Taoiseach, TΓ‘naiste, Attorney General, and Ministers for AI, Justice and Communications to discuss existing laws relating to the generation and sharing of sexual abuse images.
A European Commission spokesperson said it was analysing the Grok AI issue and there could be movement on it in the near future. The UK Prime Minister Starmer stated more action is needed to protect children on social media.
Index analysis
Concern is growing worldwide over the use of artificial intelligence tools for malicious purposes including through deepfakes, which turn genuine photos, video or audio of people into false likenesses.
Governance controls
X said it would implement technological measures to prevent Grok from allowing the editing of images of real people in revealing clothing and would geoblock the ability to generate such images in jurisdictions where it is illegal.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.