Brand safety in the AI era: When your logo shows up in ads you never approved

Somewhere online, your brand may already be appearing in contexts you never approved, endorsing questionable products, turning up at political events, or featuring on unfamiliar packaging. The content looks convincing. The logo is yours. But the brand had no role in creating it.

For years, brand safety in advertising largely meant ensuring ads did not appear alongside harmful or inappropriate content. It was a placement issue, addressed through programmatic controls, keyword blocklists and verification tools. The risk was contextual, and the response largely technical. That landscape, however, is beginning to change.

Today, the most concerning content featuring your brand may have nothing to do with your media plan. It can be created from scratch, without a brief, budget, or approval, yet appear entirely convincing.

Generative AI video tools have reached a new level of realism. Platforms such as Sora, Runway and Kling can now produce footage that appears authentic. This type of content is no longer niche; it is increasingly accessible to anyone, often free or at minimal cost. A 2024 Clarity report found that deepfake content online had grown by over 900% in three years, before the latest generation of video synthesis tools made production even easier.

The new threat landscape

What was once mainly a disinformation concern is now a marketing challenge, amplifying existing brand safety risks. Sahil Shah, CEO of Dentsu Creative Isobar, is unequivocal about the scale of what is coming. "I think this is a very serious problem, and it will only continue to grow. One reason is that India already runs on a large volume of unofficial content. There are lakhs of meme pages, fan edits, and third-party accounts that freely use branded assets without permission. Earlier, there was at least some barrier in the form of production effort, but generative AI tools have removed that barrier completely. At the same time, many of these pages are not really accountable to anyone as long as they remain within the basic guidelines of the platform."

The accountability vacuum is compounded by an audience problem that is particular to the Indian market. Shah points out that the ability to distinguish branded from unbranded content (or real from synthetic) is not evenly distributed. "A large section of Indian audiences cannot clearly distinguish between branded and unbranded content,” says Shah, “For most people, content is simply content. Perhaps the top one or two percent of audiences can identify when something is AI-generated or fake, but for the majority, the lines are blurred. This creates a situation where brands can easily appear in unsafe or misleading contexts, which is both wrong and risky."

He goes further, noting that the incentives driving synthetic content creation make the problem self-reinforcing: "Much of the fake or synthetic content being created today is designed primarily for distribution and virality, so outrageous or sensational material spreads quickly and gets consumed at scale."

The stakes extend well beyond reputational risk. "Beyond brand safety, there is also the issue of scams. We are already seeing deepfakes of celebrities being used in fake investment ads and other financial frauds. So the problem is not just about reach or viral content; it is also tied to real monetary harm."

Shradha Agarwal, Co-founder and Global CEO of Grapes Worldwide, picks up where Shah leaves off, and the anxiety she describes is visceral. "The biggest challenge of unauthorized AI content is from a perspective, if you look at it, the kind of content that people are generating — things happening from the world perspective, news perspective, and specifically when they look very realistic. Videos like this are not entertaining. It scares you because one, you don't know what is true, what is not true. Secondly, you also get scared. And then you're also wondering what will happen to your content tomorrow."

The realism of synthetic media is precisely what makes it dangerous. Traditional brand safety violations (say, a banner ad appearing on a hate speech website) were visible, traceable, and correctable. The brand was adjacent to the problem, not inside it. With AI-generated content, the brand is the content. Its logo, its visual identity, its product design, its brand characters: all of it can be remixed into a synthetic narrative that the brand never authored and cannot easily disown.

Not everyone, however, frames this as a crisis. Himanshu Goel, Co-founder and CMO of CZt, a new-age brand and communications outfit working at the intersection of culture and commerce, offers a deliberately contrarian read. "I don't see it as a threat, as user brand content has existed ever since the birth of social media. AI will just accelerate the amount and quality of such content." His prescription is to lean in rather than lock down. "If users are creating content about your brand using AI, it means your consumers care. It is an asset, not a liability."

He does acknowledge the enforcement reality: "AI watermarking is next to impossible with how advanced AI tools have become. Monitoring and automation have become powerful as well, and AI can play a role when it comes to that." It is a minority view in the current conversation, but not an entirely unreasonable one. Brands that have successfully ridden waves of user-generated content know that community energy, even when chaotic, can be channelled productively.

A legal framework catching up in real time

The advertising and legal industries are both at an inflection point. India's intellectual property regime, while largely technology-neutral in its foundational logic, was built for a world where infringement was committed by identifiable human actors operating at human speed. Generative AI breaks both those assumptions simultaneously.

Rohit Jain, Managing Partner at Singhania & Co., a full-service law firm with deep expertise in commercial litigation and IP, explains the specific legal exposure brands now face. "The rise of hyper-realistic, AI-generated ads is resulting in risk of trademark dilution in India. The unauthorized synthetic campaigns featuring logos in misleading or controversial contexts tarnish the brand's repute and also dilute distinctiveness under Section 29 of the Trade Marks Act. While pinning liability on anonymous creators or AI platforms remains legally complex, the immediate harm is to brand equity. To combat this, brands must use AI-powered monitoring tools to detect misuse early and execute rapid intermediary takedown notices to halt viral tarnishment and safeguard consumer trust."

The reference to Section 29 is significant. Trademark dilution, unlike outright infringement, doesn't require proof of consumer confusion. It only requires that the mark's distinctive character or reputation is impaired. In a world where a brand's logo can appear in a synthetic video that goes viral before the brand even knows it exists, dilution at scale becomes a very real and very fast-moving risk.

Sanjoli Jain, Counsel at Law SB, a boutique IP and technology law practice, maps out the theoretical framework with precision. "Fundamentally, intellectual property laws protect the exclusive rights of the owner. Thus, the existing intellectual property — trademark laws largely cover AI-generated advertisements that misuse brand assets, as the core principles of infringement remain technology-neutral. If an AI-generated ad uses a brand's logo, name, trade dress, or other distinctive elements in a way that causes consumer confusion or falsely suggests endorsement, it can still amount to trademark infringement, dilution, passing off, or even copyright infringement." But she is equally clear about the gap between principle and practice. "Generative AI introduces new enforcement challenges — difficulty in identifying the responsible party, the ability to generate infringing content at scale, and the increasing realism of AI-generated endorsements or advertisements that can mislead consumers."

Who bears the liability?

The question of accountability in a generative AI pipeline is not straightforward. The creator, the platform, the AI tool provider: each sits at a different point in the chain, with a different relationship to the infringing content. Sanjoli Jain lays out the hierarchy: "The creator or user of the AI tool is typically primarily liable, as they generate and circulate the content. Platforms hosting the content may benefit from limited intermediary liability if they act as neutral hosts and comply with notice-and-takedown obligations. However, they may face secondary liability if they knowingly permit, promote, or fail to remove infringing content after receiving notice. AI tool providers are generally less likely to be directly liable if the tool functions as a neutral technology. However, liability could arise if they actively facilitate infringing uses or fail to implement reasonable safeguards against misuse."

Ankit Sahni, Partner at Ajay Sahni & Associates, a law firm specializing in intellectual property and technology law, frames the enforcement challenge starkly. "Generative AI introduces a scale and speed problem that traditional enforcement mechanisms were never designed to handle. Synthetic advertisements can be created anonymously, distributed across multiple platforms simultaneously, and amplified through algorithmic promotion. While the legal principles remain applicable, enforcement becomes significantly more complex." He adds that platforms' safe harbour protections are conditional, not permanent shields. "Most intermediary liability frameworks, including those operating in India, provide safe-harbour protection only so long as the platform acts expeditiously once notified of unlawful content."

Monitoring, but at what cost?

Back in the marketing world, the practical challenge is one of resources and scale. For most brands, continuous surveillance of the open internet for synthetic misuse is not operationally feasible today. Agarwal of Grapes Worldwide is candid about the current state of play: "There is no step as a brand that you can take unless you are choosing to scan the entire internet and see for impressions about your brand to be taken and then create a legal team who constantly sues or sends emails and messages to them. It is very cumbersome today to do something like that." She notes that only a sliver of brands are even beginning to address this transparently: "Only a few brands that are very high-end and are legally authorized to do a lot of conversations around AI are mentioning the fact that this content has been created through AI. But trust me, that's only 0.01% of the brands that are doing this."

Instagram's introduction of 'Made with AI' tags represents a platform-level intervention, but as Agarwal points out, it is imperfect and already subject to workarounds. The disclosure infrastructure simply hasn't matured fast enough to match the proliferation of synthetic content.

There is also a supply-side problem that the industry is reluctant to discuss openly. Shah flags a structural vulnerability hiding in plain sight: "Branded assets such as logos, product visuals, or backshots often get widely circulated when brands work with multiple agencies, freelancers, or independent collaborators. In large agency networks or multinational setups, data privacy protocols are stronger, but in many other situations, assets are uploaded, shared, or stored rather loosely. As a result, some of this material can leak into AI training datasets, which then makes it easier for generative systems to replicate or manipulate brand imagery."

It is a reminder that the risk does not begin with a bad actor on the internet. It can begin with a Dropbox folder shared too liberally with a freelance retoucher.

Building the brand safety stack for the AI era

The emerging consensus among legal and marketing practitioners is that the response needs to be proactive, not reactive. Shah pointed out the role of platforms: "Since their business depends on advertisers, they will naturally need to strengthen their monitoring and tracking systems to ensure brand safety. Even before generative AI, brand safety was already a major concern on digital platforms, and it has traditionally been managed largely by the platforms themselves."

He also draws a pointed line for agencies themselves: "Agencies today are actively selling AI capabilities, which is understandable. However, AI should be adopted responsibly. Agencies also need to assess which categories or brands might face greater risks from AI misuse. Imagine an auto brand appearing in AI-generated accident videos that go viral for views. Even if those videos are fake, the reputational damage to the brand could be real."

Sahni advocates for an equivalent shift in how brands manage their legal posture. "Brands need to move from a reactive enforcement model to a proactive monitoring strategy. Companies should invest in continuous digital brand monitoring tools that track unauthorized use of their trademarks, logos, and product imagery across social media platforms, advertising networks, and video-sharing services." He emphasizes speed as a strategic imperative: "In an AI-driven media environment, speed of response often becomes just as critical as the legal merits of the claim."

Sanjoli Jain extends this argument to the contractual layer, noting that clear safeguards with agencies, influencers, and technology partners should restrict unauthorized AI use of brand assets and allocate liability for misuse: a point that will resonate with CMOs increasingly managing complex, multi-vendor ecosystems where the provenance of content is becoming harder to guarantee.

The industry is, in effect, being asked to treat its IP portfolio the way it once treated its media plan: with rigour, monitoring, and an enforcement mechanism. A brand's logo is no longer just a design asset. In the age of generative AI, it is a liability if unguarded — capable of appearing anywhere, in any context, doing almost anything. The brands that understand this first will be the ones that define what brand safety looks like in its next chapter.