Likeness, trademarks and harmful content

0:00 Likeness, trademarks and harm

Here's a brief that lands on social media teams every week. There's a trending moment, a famous cricketer or actor is everywhere, and someone says, let's make an AI image of them eating our biryani. It'll go viral. It might. For all the wrong reasons. In this lecture, you'll learn the risks beyond copyright: using real people's likeness, implying endorsement, misusing trademarks, and creating harmful or deceptive content. You'll learn how consent works properly, what changed in the rules in twenty twenty-five and twenty twenty-six, and a sixty second risk triage you can run on every image. By the end, you'll be able to turn a risky trend idea into a safe, effective one.

0:50 Why it matters

Why does this matter? Because these risks move fast and hit hard. A copyright question might take months to surface. A fake endorsement can trigger a takedown, a public backlash or a legal letter within hours. People have rights over their image in many countries, through publicity rights, personality rights, privacy and data protection law, passing off and defamation. Details differ, but the principle is consistent. Think of someone's face like their signature. You wouldn't sign a contract with someone else's signature. Putting their face on your ad without consent is the visual version of that. And in some markets, including the Gulf, laws on privacy, defamation and cybercrime can carry serious penalties.

1:39 Four risk areas

Let's break it into four risk areas. First, likeness: generating or using images of real, identifiable people, whether celebrities, influencers, politicians or private individuals, without consent. Second, implied endorsement: showing someone apparently using or recommending your product, which can mislead consumers and infringe their rights. Third, trademarks and brand assets: other brands' logos, packaging, characters and distinctive products, including pseudo-logos the model invents that resemble real ones. And fourth, harmful and deceptive content: deepfakes, sexualized or humiliating imagery, fake news-like scenes, and content that exploits tragedies or sensitive topics. Here's the key idea. Each area needs a different response, from consent to redesign to simply not publishing.

2:26 What changed

Here's what changed recently. In the European Union, the AI Act's transparency rules apply from August twenty twenty-six. People who use AI to create deepfakes, meaning realistic content that resembles real people, places or events and would falsely appear authentic, must disclose that it's AI-generated or manipulated, with lighter treatment for clearly artistic or satirical work. In the United States, the TAKE IT DOWN Act of twenty twenty-five made publishing non-consensual intimate imagery, including AI-generated forgeries, a federal offense, and requires platforms to remove it on request. Several US states also protect voice and likeness against AI replicas. And platforms have tightened policies and added removal tools for people who are simulated without consent.

3:16 Consent done properly

Now consent done properly. If you use a real person's likeness in synthetic content, like a creator's AI avatar or a founder in an AI-generated campaign, you need written consent with a clear scope. Which uses, channels and markets. How long. Their right to review outputs. Payment terms where relevant. How they can withdraw consent and what happens to existing content. Confirmation that the tool allows this use. And a disclosure plan. Verbal okays in a group chat don't count. And remember private individuals too. Customers, employees and members of the public have privacy and data protection rights. Get consent before uploading their photos to AI tools, let alone altering them.

4:04 Worked example 1: a coach on a fake summit

Worked example one, simple. A fitness coach wants to post an AI image of herself on a mountain summit she's never climbed, to promote a mindset program. It's her own likeness, so consent isn't the issue. But run the triage. Could it mislead about an event? Yes, followers might believe she actually climbed it. So either she labels it clearly as an AI-generated illustration, using the platform's label and a caption note, or she chooses a clearly stylized, illustrated version instead. She goes with the illustration. It's still eye-catching, it can't be mistaken for a real photo, and it fits her brand style kit.

4:49 Worked example 2: the biryani trend

Worked example two, the biryani trend, a business scenario with illustrative details. A restaurant in Karachi wants to jump on a trend with an AI image of a famous cricketer tasting its new biryani. Let's triage. Real person identifiable? Yes. Consent? No. Stop there. Implied endorsement? Yes. Platform risk? High, likely removal. So what's the better idea? First, the authentic version. Photograph staff and regular customers, with signed consent, trying the dish, with genuine reactions. Second, the playful version. Create an illustrated cartoon chef character who reacts to the biryani in the trend's format. It's ownable, reusable and safe. The restaurant still rides the trend, without borrowing anyone's face.

5:36 Watch me do it: 60-second triage

Watch me do it. I'll run the sixty second triage on a real-world style ad. The image shows a smiling family in a car, with a generated city skyline behind them. Question one, is a real person identifiable? The family are AI-generated, not real people. No. Question two, does it imply endorsement? No. Question three, any third-party logos or landmarks with rights? Let me zoom. There's a pseudo-logo on a building that looks a lot like a real bank. Yes. So I inpaint it out. Question four, could it mislead about a product or event? The car is a real photo composite, and the claims are accurate. No. Question five, sensitive topic? No. Risk level: low, after the fix.

6:28 Off the table

Let's talk about the harmful content zone, because some things are simply off the table. Non-consensual intimate or sexualized imagery of real people is unlawful in many countries and banned on every major platform. Fake news-like images of real events, disasters or conflicts can spread misinformation and hurt people. Content exploiting tragedies, children, or religious symbols can cause deep offense, and in some markets, legal consequences. Political and social issue ads have stricter rules on most platforms. If something goes wrong, act fast. Take the content down, tell the affected person or brand, document what happened, and fix the process that let it through.

7:13 Common mistakes

Common mistakes. Using celebrity or influencer likeness because it's trending. Believing that a parody label makes any use acceptable. Uploading customers' photos to AI tools without consent. Missing pseudo-logos that resemble real brands. Using real landmarks or other brands' packaging in ways that suggest affiliation. Treating sensitive topics casually. And having no escalation path, so a junior designer has to make a legal call alone at midnight. Define your risk levels. Low means publish after QA. Medium means a second reviewer. High means legal or client sign-off. Then nobody has to guess.

7:53 Recap and try this now

Let's recap. Beyond copyright, watch four risk areas: likeness, implied endorsement, trademarks, and harmful or deceptive content. Get written, scoped consent for any real person, including private individuals. Know the new rules: EU deepfake disclosure from August twenty twenty-six, the US TAKE IT DOWN Act, state likeness laws and stricter platform policies. Run the sixty second triage on every image and escalate by risk level. Try this now. Take ten AI image ideas for a brand, run the triage template from the lesson on each, and for every medium or high risk idea, write the review, consent or alternative approach you'd use. You'll be surprised how often the safe version is also the more creative one.