When AI-related problems surface in advertising, the first question is usually about the law. But for the people managing production budgets and agency relationships, the practical question is different: if something goes wrong with AI-generated content, does your current setup give you the visibility and the contractual standing to manage the situation?

This is the final article in a three-part series on the AI legal landscape for global advertisers. The first article examined whether AI-generated campaign assets can be legally protected. The second looked at what contracts need to say. This one covers what has already happened in the courts and with regulators, and what it means for how you manage your production partnerships.

The largest UK AI copyright case: Getty Images v. Stability AI

Getty Images v. Stability AI [2025] EWHC 2863 (Ch), UK High Court, November 2025

Getty Images sued Stability AI alleging that millions of Getty's copyrighted photographs were used without permission to train the Stable Diffusion image generation model. The case became the first major AI copyright case to reach judgment in the UK. However, Getty accepted during proceedings that the training did not occur in the UK. Because the acts of copying could not be demonstrated within UK jurisdiction, the primary copyright claims were abandoned. The court ruled that the AI model's learned parameters (its "weights") are not a "copy" of the training images. Getty won limited trademark findings related to watermarks appearing in some outputs, but the core copyright question remains unresolved.

Roch noted that the jurisdictional issue will be significant going forward. If AI models are trained outside the UK, and most are, UK copyright holders face a substantial hurdle in bringing domestic claims. A reference to the European Court of Justice on whether AI outputs can be protected by copyright is pending but is unlikely to yield a decision before 2027.

The unintentional brand incorporation risk

Roch flagged a risk that is worth paying attention to from a production management standpoint. AI tools trained on advertising content from multiple brands can inadvertently reproduce elements like imagery, design patterns, or visual language from one brand's campaigns in content generated for another.

His example: if an AI tool trained on KFC advertising produces imagery for McDonald's that contains recognisable KFC visual elements, neither party may realise it until the content is in market. This is difficult to detect without deliberate review, and if it does happen, liability will depend on what the contracts say. If your agency agreement doesn't address AI-generated content and third-party rights, you may not have the standing to hold the agency accountable for something neither of you saw coming.

Advertising standards: existing rules apply to AI content

AI does not create a separate regulatory regime for advertising. The existing rules on misleading content, objectification, and consumer harm apply regardless of how the content was produced. Roch identified two ways in which advertising standards authorities are already engaging with AI.

AI-generated imagery and misleading claims

If a cosmetic brand uses AI to generate an image showing the effects of a tanning product, and adds a disclaimer that the image is AI-generated, the advertising authority may find that the disclaimer itself makes the claim more confusing. The consumer is left wondering whether the product actually produces these results. A disclosure intended to increase transparency can itself trigger a ruling.

AI tools whose advertising implies harmful capabilities

ASA Ruling: PixVideo, AI Video Maker, UK ASA, March 2026

The UK Advertising Standards Authority banned a YouTube ad for PixVideo after receiving eight complaints. The ad showed before-and-after images of a young woman with text reading "Erase anything," strongly implying the app could digitally remove a woman's clothing. The company stated that its app does not permit users to create sexually explicit content. The ASA ruled based on the impression the ad created, not the technical capabilities of the product.

The principle from both examples is the same. Regulators assess the impression an ad creates. AI does not exempt advertisers from existing rules, and in some cases, AI-generated content may attract greater scrutiny. For production teams reviewing agency work, this means that AI-generated content needs the same level of compliance review as traditionally produced content, and possibly more.

Emotion recognition: an emerging and high-risk use case

Roch identified this as one of the most significant areas to watch at the intersection of AI and advertising.

Network Rail Emotion Detection Trial, UK, revealed June 2024

Network Rail conducted a two-year trial using Amazon Rekognition AI cameras at eight major stations, including London Euston, London Waterloo, Manchester Piccadilly, and Glasgow Central. The cameras analysed passengers' facial expressions to estimate emotions, age, and gender. Network Rail stated the data could be used to "measure satisfaction" and "maximise advertising and retail revenue." The trial was revealed through a freedom of information request by Big Brother Watch, which filed a complaint with the Information Commissioner's Office. The emotion and demographic analysis portion was subsequently ended.

Under the EU AI Act, this type of emotion recognition would be classified as a high-risk AI system. Roch expects similar deployments to emerge in other environments like fast food restaurants, retail spaces, and digital signage, where the commercial incentive is strong but the regulatory framework may not be in place.

These tools are sometimes proposed by agencies or vendors as consumer insight or measurement capabilities making them relevant to procurement and marketing teams. If a vendor is proposing emotion recognition or demographic profiling as part of a campaign measurement package, the regulatory and reputational exposure that comes with it may not be obvious from the proposal. It is worth asking specifically what technology is involved and what data is being collected.

What advertisers should be doing now

The legal landscape around AI in advertising production will not resolve quickly in any jurisdiction. Based on our conversations with both Roch Glowacki and Chris Mammen, we recommend that advertisers take the following steps.

1. Audit contracts across all markets

for AI-specific provisions covering ownership, data usage, tool disclosure, and liability allocation. Calibrate to the actual risk of each AI deployment rather than applying a blanket clause.

2. Require disclosure of AI use from all production partners.

Ask not just what tools are being used, but for what purpose and with what data. This is also directly relevant to evaluating whether the production costs you are being charged are appropriate for the work being done.

3. Distinguish between content that needs protection and content that does not.

Long-lived brand assets face real commercial risk from the absence of copyright protection. Ephemeral social content may not require the same level of concern. Both your production costs and your contract terms should reflect which category the work falls into.

4. Review insurance coverages

to determine whether existing policies cover AI-related brand damage, IP infringement through AI outputs, and right of publicity claims. Cyber liability policies designed around data breach exposure may not adequately cover this type of situation.

5. Evaluate whether your agency's use of AI is reflected in what you're paying.

If AI tools are reducing the time, labour, or resources required to produce content, the production costs should reflect that. If they don't, it is worth having the conversation.

6. Strengthen RFP processes

to capture AI usage at the proposal stage. Specific questions about tools, purpose, data, and governance are more likely to surface useful information than broad ones. And understanding how AI is being used in a proposed production workflow is directly relevant to evaluating the proposed pricing.

7. Be cautious about vendor proposals involving consumer analysis tools.

Emotion recognition, demographic profiling, and real-time behavioural analysis carry serious regulatory and reputational risk. Ask specifically what technology is involved and what data is being collected before approving any proposal that includes these capabilities.

8. Monitor developments in each market

where your advertising appears. U.S. state legislation, the UK's AI and IP consultation, the EU AI Act's implementation, and the pending European Court of Justice reference are all moving on different tracks. Quarterly check-ins with legal counsel are your key markets is a practical way to keep pace.

The decisions being made now about how AI is used, disclosed, and governed in the production process will determine your exposure when the legal landscape does stabilise. They will also determine whether you are getting fair value for the production work being done on your behalf.

About this series

This three-part series with Roch Glowacki builds on BBS's earlier conversation with Chris Mammen of Womble Bond Dickinson, which covered the U.S. perspective on AI liability in advertising production. Together, the two conversations reinforce a consistent finding. The legal framework will take years to settle, and advertisers need contractual protections, internal governance, and informed oversight of their production relationships rather than waiting for the law to catch up.

Read the full U.S. perspective: Who Owns Your AI Advertising Mistakes?

Roch Glowacki is a commercial contracts lawyer at Lewis Silkin LLP, advising advertisers and agencies on AI governance, commercial contracts, and regulatory compliance. BBS Worldwide is an independent advertising production consulting firm that helps advertisers manage production costs, protect rights, and strengthen agency relationships globally.

This article is for informational purposes only and does not constitute legal advice. The legal perspectives shared here reflect general observations about the current state of the law and should not be relied upon as guidance for specific legal decisions. Readers should consult qualified legal counsel for advice on their particular circumstances.