Generative AI: Navigating the ethical and legal minefield

upGrad and Zomato recently used images/videos of celebrities generated by AI programs in ads. Both have taken down the ads from their social media handles.

Generative AI and advertising may seem like a match made in heaven; but not without some caveats.

When a computer spits out visually striking images within minutes of receiving a basic textual prompt, it feels magical. The creative advertising community has been among the few sectors to explore this fascinating new tool.   

For some agency folks, image generators like Midjourney and Stable Diffusion have come in handy to automate the initial stages of conceptualising a campaign — simple visualisations, and storyboarding — and others have directly used images created by these programs in campaigns with minimal alterations.

But as it happens with any new technology, and especially one that is as path-breaking as generative AI (GenAI), the boundaries of what can be done and what shouldn’t be done are just about being defined. There are several ethical and legal implications of using images conjured up by GenAI for commercial purposes. 

The intellectual property question is one that is already being litigated in the US. A group of artists has sued multiple generative AI platforms, including Midjourney and Stable Diffusion, for using their artworks without consent or compensation to train the platforms’ algorithms. 

In India, brands are pushing the envelope to explore the limits of GenAI. Recently, ed-tech brand upGrad used the likeness of Google chief Sundar Pichai in an advertisement. Zomato, a food delivery platform, used the likeness of football player Lionel Messi, actor Leonardo DiCaprio and Tesla chief Elon Musk in a video. 

upGrad faced severe backlash and has since removed the Sundar Pichai visual from its social media channels. Zomato too has taken down the AI-generated video from Instagram. (afaqs! contacted both brands but did not hear from either until the time of publication)

By the Law

Legal experts that afaqs! spoke to stated in no uncertain terms that brands cannot use the likeness of celebrities without their consent, even if the images or videos have been created using generative AI. The advertising community’s self-regulatory body is observing the use of GenAI by brands with a keen eye. 

Lavin Hirani, Partner and Media & Entertainment practice head, MDP & Partners, says that at present, there is no piece of law that explicitly grants or defines personality rights in India, however, several judgements of Indian and foreign courts have applied principles of common law to hold infringers liable for their actions. 

He cites examples of actors like Amitabh Bachchan and Katherine Heigl where brands used certain characteristics of the actor and their images without consent and Indian and foreign courts held the brands liable for their infringement of the Right to Publicity and Image Rights. 

The only times that using a celebrity’s likeness can possibly avoid violating publicity laws is if “the likeness is completely transformed or is used for parody, that too in small measures”, says Mathew Chacko, Partner, Spice Route Legal. “At the same time creating a parody or joke based on the celebrity’s image or attributes in a disguised manner to subvert the prevailing laws is also prohibited,” he adds.  

Using GenAI to produce convincing and accurate representations of a celebrity's appearance, voice, mannerisms and behaviours is far easier and more common now. So far, brands like Cadbury Dairy Milk and Ageas Federal Life Insurance have used “deep fakes” with the consent of a celebrity (Shah Rukh Khan and Sachin Tendulkar respectively) to generate remarkable campaigns. It is unclear if upGrad and Zomato sought consent to use the images of celebrities in their ads.  

Due Diligence Lacking 

Generative AI has caught the fancy of both brands and agencies who want to be the first to use the technology to create a campaign that will wow both consumers and the community at large. In the process, teams creating the ads might end up sacrificing due diligence.  

“Content and ads are being churned out at high speed that I feel robust due diligence is not being done internally on some of the work being put out. This is why we see violations even for companies where you expect to have sound processes and enough checks and balances — especially on social media ads,” says Manisha Kapoor, CEO, Advertising Standards Council India. 

She says new technology brings a new dimension to content and advertising but old rules still apply. The problem arises when executives forget that “the principles of advertising remain the same irrespective of the technology used to make the ad.” 

When in doubt… 

Agencies have been toying around with image generators to create refreshing visuals for brands. Schbang recently used generative AI tools to make a world of JimJam biscuits. Its co-founder and CEO Akshay Gurnani says that one way to avoid legal tangles is “to avoid including references to preexisting copyrighted works or characters in prompts.” He further adds “do not upload preexisting copyrighted images unless you own the copyright or a valid license in those images.” 

The magic of image generation is all in the prompt and how it is framed. A prompt that asks the program to create a bedroom with a bed, a chest of drawers, and a lampshade in the IKEA style, will do just that. Using that image which is modelled on an IKEA bedroom for another furniture brand is bound to trigger a whole host of violations. 

Mitesh Kothari, co-founder and CCO, White Rivers Media, says brands that are showing interest in using GenAI images are also coming to the agency with queries about the legalities of using such content. 

White Rivers Media recently generated very realistic images of the cast of Fast X visiting iconic Indian travel destinations — Jason Mamoa in Goa, Vin Diesel at the Taj Mahal in Agra, and Jason Statham on a boat on Dal Lake in Kashmir. This was part of the marketing communications for the movie’s launch in India.  

Using the likeness of celebrities is just one aspect of the moral and legal dilemma surrounding the use of image generators. The other is that of imitating the artistic styles of creators. Kothari says that brands are mostly in the clear when they use their own brand assets and brand attributes to generate an image. 

Essentially, the brand should steer clear of asking the program to generate an image based on existing works of art and artists like Picasso or Van Gogh. Kothari however wonders, “what if a human artist took inspiration from an art style and created something in that style for a brand?” Why would that be above board and the other a violation of copyrights and trademarks, etc?  

Mithun Mukherjee, ECD at FCB Kinnect, has a simple approach towards AI — “either use it to make a point or multiply creativity”. FCB Kinnect used an image generator to create striking visuals of women playing Holi, a festival that women typically avoid playing in public because of the fear of being molested. The campaign for vivo made the point that such images can only be generated using AI because they are not real images of real women enjoying the festival.

View this post on Instagram

A post shared by vivo India (@vivo_india)

Mukherjee cites the example of Coca-Cola’s campaign as part of which the beverage brand invited artists to generate original artwork with creative assets from the Coca‑Cola archives. Co-creating content opens up new ways for people to engage with the brand, thereby multiplying creativity and making something that was hitherto impossible, possible.  

Training the Humans running AI

Finally, it all comes down to education and awareness creation. “It is important for everyone who is involved in releasing an ad to be aware of what is right and wrong. Creativity will try and push boundaries — which is why educating the junior-most person on the team is very essential. That is what is missing in the rush to put out something,” says Kapoor.  

Mukherjee recognises that using GenAI for advertising is still in its nascency in India, which is why people are still fiddling around to understand what they can do with the program. “Going into what is right and wrong is the next phase that we will step into when everyone knows how to use this technology better,” he says. For now, Mukherjee is focussing less on the ‘dont’s’ because he wants the people in his team to experiment as much as possible. “Telling people what they should not try could inhibit their imagination,” he says. 

For Kapoor, the consumer’s interest is primary. ASCI may consider clarifying rules if the regulatory body sees more instances of violations and misleading ads. “The ultimate question about ads is whether or not the content is misleading. And that content could be made by anyone, human or AI.”