A viral challenge with costumes and AI images of the 'Cat in the Hat' threatens US schools and has already led to teen arrests

By The Guardian · September 13, 2026.

Several teenagers have been arrested or charged in the United States for taking part in a viral challenge that repurposes the Cat in the Hat —the Dr. Seuss character, in the unsettling version inspired by the 2003 Mike Myers film— to simulate threats against schools. According to the Hutchinson (Kansas) police, a 15-year-old student at Hutchinson High School created a TikTok account with messages that police classified as a criminal threat; he was charged on two counts over it. In Worth County (Georgia), two high school students face charges and disciplinary sanctions from the school district itself. In Grand Junction (Colorado), police stepped up their presence at two schools 'as a precaution', according to their statement, after a post from the challenge circulated, ending with the arrest of a 14-year-old.

The format combines two distinct ingredients: AI-generated photos and videos, and real people dressed up as the character, prowling at night beside houses, streets or schools. Some posts include text messages aimed at specific towns or schools; the most extreme versions name individuals or what are presented as target lists. As the BBC reconstructs it, the trend started in Ireland and the United Kingdom in August 2026 before quickly jumping to parts of the United States and Canada, via Snapchat.

The authorities' response is the fact that really matters here: although for a good share of users this is a joke or a homemade horror format, the Hutchinson police said it without qualification —'regardless of whether a post is conceived as a joke, it is illegal to make others, particularly students, fear for their safety'—. It is the same logic we already saw in another recent episode, when more than 400 people were arrested at a July 4 gathering organized through TikTok in California: a social media format designed to go viral collides with a legal framework that does not distinguish playful intent from the real effect on the victim.

What is relevant from a technological standpoint is not that generative AI has created a terrifying character —humans have been doing that with costumes forever—, but that it has driven to zero the cost and the skill needed to produce a convincing harassment or threat video. A teenager with no technical ability can generate in minutes a believable night-time scene with a recognizable character prowling outside a school with real names superimposed. That changes the risk equation: a few years ago, staging a threat like this would have required a camera, editing and a certain effort that acted as a natural filter; now that filter has disappeared, and the result is that minors run afoul of the law for producing content whose credibility they themselves do not gauge.

That, at bottom, is the short-term friction we have been pointing to in other areas of mass AI adoption: the technology democratizes content-production capabilities faster than institutions —schools, police forces, platforms— manage to set clear rules on their use. This is not an existential threat or a national security matter; it is a problem of digital literacy and platform moderation that is solved, in the medium term, through education about legal consequences and through those same AI tools applied to detecting and containing this kind of content at scale, instead of leaving each case to depend on a school or a police station catching it in time. In the meantime, the authorities' recommendation —report it to a trusted adult instead of forwarding the video— is, for now, the only patch available.