The ASA Found These Ads Without a Complaint: What AI-Driven Enforcement Means for UK Advertisers

If you run paid social in the UK, you may still work on the assumption that a borderline ad only becomes a problem once somebody complains. On 16 September 2026 the Advertising Standards Authority (ASA) published five rulings that make that assumption hard to hold. They concern six Meta ads for AI apps, and none of them started with a complaint. Each was found by the ASA's own Active Ad Monitoring system, which, in the regulator's words, uses AI to proactively search for online ads that might break the rules.

What the ASA ruled

The rulings name Animcha Ltd t/a tyan.ai, an AI companion app; Guangxi Baise Tianyi Agricultural Development Co Ltd t/a Nexaipic, an AI portrait tool with two ads; Persons unknown t/a KH31 DD22 and Sinosea Technology PTE Ltd t/a PictoPop, both image-to-video apps; and Rusto AI-AI Photo Toolbox t/a Rusto AI, an image generator. All ran as paid ads on Meta, and the ads were seen between 10 March and 2 April 2026.

Four of the apps were advertised as turning a photograph into a sexualised video; the fifth, tyan.ai, as a customisable "AI girlfriend". The ASA upheld every case. It found the ads irresponsible, likely to cause serious offence and to include a harmful gender stereotype, because they presented women as sexual objects, breaching CAP Code rules 1.3, 4.1 and 4.9. For the four photo-to-video apps it also found the ads condoned manipulating women's images into sexually explicit content without consent. The tyan.ai ad drew a second finding, under rules 1.3, 4.1 and 4.8: the ASA judged that it portrayed a young woman who appeared to be under 18 in a sexual way, citing her styling and a setting that looked like a child's bedroom.

None of the advertisers answered the ASA's enquiries, and each ruling records that silence as a breach of rule 1.7, under which any unreasonable delay in responding will normally be considered a breach of the Code. Four also cite rule 1.7.1, because a party identified only as persons unknown gave no name or business address. Meta confirmed the ads had run on its platform, said they breached its advertising guidelines and standards, and said they had been removed before the ASA contacted it. The rulings were published regardless. The ads must not appear again in the form investigated, and every case was referred to CAP's Compliance team.

The law moved while the cases were open. Section 99 of the Crime and Policing Act 2026 inserted a new offence into the Sexual Offences Act 2003 of making, adapting or supplying a tool for generating purported intimate images, the category that includes so-called nudification apps. It came into force on 29 June 2026 and applies in England and Wales.

How the ASA finds ads now

The ASA describes Active Ad Monitoring in three parts. It captures ads from social media, search, websites and influencer marketing. AI models filter that volume for ads likely to be relevant to a compliance problem. ASA staff then review what is flagged and decide whether to act. The regulator says the system processed more than 60 million ads in 2025. In November 2023, when it expected to put more than three million ads through the system that year, it announced a five-year strategy for 2024 to 2028 to invest more in proactive work than in reactive complaints casework.

The ASA is explicit that the models do not decide; its experts review what is flagged and decide whether to act. The model's job is to put a likely breach in front of a person. A low-spend ad shown to a narrow audience that would never complain can still be captured, surfaced and ruled on. Obscurity was never a defence under the Code, but for years it worked as a practical one. At the scale the ASA now reports, it no longer does.

The ASA says the system supported more than 30 regulatory projects in 2025, including work on environmental claims, gambling, alcohol, prescription-only weight-loss medicines and cosmetic surgery offered from outside the UK. If you advertise in any of those categories, plan on the basis that your ads may already have been captured.

Five assumptions the rulings retire

The CAP Code places primary responsibility on marketers, but states that agencies, publishers and other service suppliers also accept an obligation to abide by it. In three of the five rulings the ASA describes the ad as run by a different party from the app company, such as the tyan.ai ad "by Scarlett", and it addressed its findings to both. For an agency running UK campaigns, a ruling against a client is also a question about its own review.

Ignoring the regulator has costs beyond the ruling. For advertisers unwilling or unable to comply, the sanctions include CAP Ad Alerts advising media owners to withhold services such as ad space, the ASA's own paid-search ads highlighting an advertiser's name and non-compliance, and, for those who persistently break the rules, referral to statutory bodies such as Trading Standards.

What to check this quarter

None of these ads was borderline, and most brands will never run anything close to them. The lesson for mainstream advertisers is procedural. The regulator now looks at what you run whether or not anyone objects, so your own review has to happen before launch and cover everything that actually serves.

1. Review the variants that serve, not only the master creative. Dynamic and platform-assembled ads recombine images, headlines and text, and each combination is an ad the ASA can capture.
2. Treat product demos as claims. If you sell an AI tool, the sample output in the ad is the ad. These rulings turned on what the apps were shown producing and on lines such as "turn photos into hot motion".
3. Judge synthetic people as if they were real. Age, styling, setting and pose are assessed on how they appear. If a generated person could read as under 18, change the image.
4. Put a named owner on ASA correspondence, with a deadline. A legitimate brand that misses one adds a Code breach to whatever else is at issue.
5. Use CAP's Copy Advice team for close calls. It reviews concepts, imagery and copy, with a free standard service and paid faster options, but its advice is not binding on the ASA, so treat it as a second opinion rather than clearance.
6. Keep a record of what was approved, by whom, when, and with which targeting. The ads here were seen in March and April and ruled on in September, so you need to be able to reconstruct a campaign months later.

Disclosure is a separate question. Labelling AI-made ad creative tells people how an ad was made. These rulings are about what an ad shows and says, and a label would not have saved any of them.