A controversial AI-generated anti-drug video, designed to warn young people about the dangers of substance abuse, has been condemned by UK regulators as ‘dangerously alluring’ for glamorising the very behaviour it seeks to prevent. The Advertising Standards Authority and the Home Office have called for an urgent review of the campaign, which was produced by a tech startup using generative adversarial networks to create hyper-realistic imagery.
The video, which has been viewed millions of times across social media platforms, depicts a series of vivid scenes: teenagers at a party, the glow of neon lights, a pill dissolving into a drink, then a rapid descent into addiction, homelessness and overdose. The problem, critics say, lies in the aesthetics. The AI has rendered the early scenes with such cinematic beauty – saturated colours, perfect lighting, fluid motion – that they inadvertently make drug use look appealing.
‘This is the Black Mirror paradox of AI ethics in real time,’ said Dr. Alisha Kapoor, a digital ethicist at the London School of Economics. ‘The technology is so good at creating compelling narratives that it loses sight of the message. Young viewers are more likely to share the video for its visual spectacle than to internalise the warning.’
The startup behind the video, NeuroVision AI, defended its work. ‘We used cutting-edge models to ensure the content resonated with a generation raised on Instagram and TikTok,’ said founder Marcus Cole. ‘If the imagery is striking, that’s the point – we need to cut through the noise. The ending is unambiguous in its horror.’
But the data tells a different story. Early analytics show that viewers are skipping the final third of the video, with the most replayed segments being the party scenes. Engagement metrics indicate that the video is being used in ‘mood board’ playlists and remixed with upbeat music. The ASA has received 140 complaints, with many parents reporting that their children asked for ‘the pretty drug video’ to be played again.
This incident highlights a growing concern around AI’s role in public health messaging. Traditional anti-drug campaigns from the 1990s relied on stark, low-budget realism: blurred camera work, washed-out colours, and blunt dialogue. Today’s generative models, however, are trained on vast datasets of cinematic history and can produce visuals that rival Hollywood productions. The result is a medium that can seduce even as it warns.
‘We are entering an era of synthetic persuasion,’ said Julian Vane, Technology & Innovation Lead. ‘Every new algorithm carries a responsibility we are still learning to design for. This video is a case study in the law of unintended consequences. The user experience of society is at stake: we cannot have our public health tools look more alluring than the behaviour they are trying to prevent.’
The Home Office has suspended further distribution of the video pending the review, which will examine whether AI-generated content should be subject to stricter pre-publication testing. Proposed measures include mandatory ethical impact assessments for any government-funded AI creative work, and a requirement that generated content includes visible watermarks or disclaimers.
For now, the video serves as a cautionary tale. The very power of AI to captivate and engage has become its liability. In trying to save young people from peril, it may have inadvertently placed them closer to the edge. The review is expected to report within two weeks.
