---
title: "Likeness, trademarks and harmful content"
description: "People, brands and harm Beyond copyright, AI imagery raises other legal and ethical risks: using real people's likenesses, misusing trademarks, and…"
url: https://optimizeall.com/learn/ai-image-generation-and-design/likeness-trademarks-and-harm
updated: 2026-10-05
---

AI Image Generation and Design · Copyright, likeness and disclosure · lesson 14 of 18 · 7 min

# Likeness, trademarks and harmful content

## People, brands and harm

Beyond copyright, AI imagery raises other legal and ethical risks: using real people's likenesses, misusing trademarks, and generating content that is deceptive, harmful or offensive. These risks can damage reputations and relationships fast — and some are criminal offences in certain countries.

## Likeness and personality rights

Many jurisdictions protect people against unauthorized commercial use of their name, image or likeness, through rights of publicity, personality rights, privacy law, data-protection law, passing off or defamation. Details differ widely, but the principle is consistent:

- **Do not generate or use images of real, identifiable people** (celebrities, influencers, politicians, private individuals) in marketing without their consent.
- **Implied endorsement** is a particular risk: showing a well-known person apparently using your product can mislead consumers and infringe their rights.
- **Deepfakes and sexualized or humiliating imagery** of real people can be unlawful in many countries and violate every major platform's policies. Several jurisdictions have introduced specific offences for non-consensual intimate deepfakes.
- **Private individuals** — customers, employees, members of the public — have privacy and data-protection rights over their images. Get consent before uploading their photos to AI tools or altering their likeness.

In Gulf countries, laws on privacy, defamation, cybercrime and public morals can be strict, and publishing images of people without consent — or content deemed offensive — can carry serious penalties. Pakistan's cybercrime law also covers certain misuse of images. Check local laws and err on the side of consent.

## Consent done properly

For synthetic content of a real person (for example, a creator's AI avatar, or a spokesperson's likeness used in ads):

```
Likeness consent checklist
[ ] Written consent from the person
[ ] Clear scope: which uses, channels, markets, duration
[ ] Right to review/approve outputs
[ ] Payment/compensation terms where relevant
[ ] How to withdraw consent and what happens to existing content
[ ] Tool compliance: the tool permits this use and has consent processes
[ ] Disclosure plan for synthetic content
```

## Trademarks and brand assets

- Avoid generating other brands' logos, packaging or distinctive products, which can imply affiliation or infringe trademarks.
- AI can accidentally generate pseudo-logos that resemble real brands — check backgrounds, clothing and products.
- Don't create "parody" ads of competitors without legal advice; comparative advertising has strict rules.

## Harmful, misleading and sensitive content

Most tools have content policies, and platforms have community guidelines and ad policies. Beyond compliance, professional ethics require avoiding:

- **Misinformation:** realistic images of events that did not happen, especially news, disasters, elections or conflicts.
- **Fake evidence:** images presented as proof (fake screenshots, fake receipts, fake reviews).
- **Stereotypes and demeaning depictions** of groups based on ethnicity, religion, gender, disability or other characteristics.
- **Religious and cultural sensitivity:** depictions of religious figures, sacred texts or sites may be deeply offensive or unlawful in some markets.
- **Children:** extreme care; many platforms prohibit certain depictions, and some content is illegal everywhere.

## A risk triage for every AI image

```
Low risk:    Abstract textures, clearly stylized illustrations, landscapes, generic objects
Medium risk: Realistic generic people, realistic scenes, product composites, cultural motifs
High risk:   Real people, public figures, children, news/events, religious content,
             health/finance claims, other brands' assets
→ Medium: extra review. High: consent/legal review or don't proceed.
```

## Worked example: a trending celebrity edit

A creator wants to post an AI image of a famous footballer wearing the creator's merch to ride a trending topic.

Analysis: identifiable real person, commercial context, implied endorsement, no consent → high risk under likeness laws, advertising rules and platform policies. Alternative: create a clearly original illustrated character, or reach out to the athlete's management for a paid partnership.

## Common mistakes

- "It's just a joke/meme" — commercial accounts face stricter scrutiny.
- Using customers' photos in AI tools without consent.
- Missing pseudo-logos in backgrounds.
- Generating realistic images of events for engagement.

## When something goes wrong

Mistakes happen, so plan your response. If you discover that a published image contains a real person's likeness without consent, another brand's logo, or a misleading or offensive element, remove or edit it promptly, inform the client, and document what happened and how it will be prevented. If someone complains, respond respectfully and quickly. Updating your style system, risk triage and review steps after each incident turns a mistake into a stronger process.

## 2025–2026 update: the rules got sharper

- **EU AI Act transparency obligations (from August 2, 2026).** Deployers who use AI to generate or manipulate image, audio or video content that resembles real people, objects, places or events and would falsely appear authentic ("deepfakes") must disclose that it is artificially generated or manipulated, with lighter treatment for evidently artistic, creative or satirical work. The European Commission has been developing a code of practice on marking and labeling AI-generated content to support these rules.
- **United States.** The TAKE IT DOWN Act (2025) made publishing non-consensual intimate imagery, including AI-generated "digital forgeries", a federal offense and requires covered platforms to remove it on request. Several states also protect voice and likeness against unauthorized AI replicas.
- **Platforms.** Major platforms prohibit deceptive synthetic depictions of real people and label realistic AI content; several offer tools for people to request removal of AI content that simulates them.

## Hands-on: the pre-publication risk triage (60 seconds per image)

```text
RISK TRIAGE — asset: ______
Real person identifiable (face, voice, name, distinctive features)?  Y / N
  → Y: written consent on file? scope covers this use? → if not, STOP
Implies endorsement or affiliation by a person or brand?             Y / N
  → Y: permission on file? otherwise redesign
Third-party logos, packaging, characters, landmarks with rights?     Y / N
  → Y: remove, blur, or obtain permission
Could it mislead about a product, result, event or news?             Y / N
  → Y: real imagery instead; AI only where it cannot mislead
Sensitive topic (health, finance, politics, religion, children, tragedy)? Y / N
  → Y: senior review + platform/ad policy check + disclosure
Risk level: LOW (publish after QA) / MEDIUM (second reviewer) / HIGH (legal or client sign-off)
```

## Before/after: a trend that almost went wrong

| | Before | After |
|---|---|---|
| Idea | A viral "celebrity tries our biryani" AI image for a Karachi restaurant | Staff and regular customers (with signed consent) photographed trying the new dish; an illustrated cartoon chef for the playful version |
| Risks | Likeness and implied endorsement; platform removal; possible legal action | None of those; the illustrated version is ownable |
| Result | Would likely be removed, and could damage reputation | Authentic content plus a reusable brand character |

## Summary

Never use real people's likenesses without consent, avoid other brands' assets, don't create misleading or harmful imagery, respect cultural and religious sensitivities, and triage every image by risk level.

## Video lecture: Likeness, trademarks and harmful content

Lecture coming soon · 11 chapters · about 9 minutes. Read the full transcript below.

1. Likeness, trademarks and harm
2. Why it matters
3. Four risk areas
4. What changed
5. Consent done properly
6. Worked example 1: a coach on a fake summit
7. Worked example 2: the biryani trend
8. Watch me do it: 60-second triage
9. Off the table
10. Common mistakes
11. Recap and try this now

## Lecture transcript

### 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.

### 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.

### 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.

### 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.

### 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.

### 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.

### 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.

### 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.

### 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.

### 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.

### 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.

## Key takeaways

- Beyond copyright, manage likeness, implied endorsement, trademarks and harmful or deceptive content.
- Use real people's likeness only with written, scoped consent — including customers and employees.
- EU AI Act deepfake disclosure applies from August 2026; the US TAKE IT DOWN Act (2025) targets non-consensual intimate imagery, including AI forgeries.
- Check outputs for pseudo-logos, third-party packaging and landmarks that imply affiliation.
- Run a 60-second triage on every image and escalate by risk level (QA, second reviewer, legal/client).

## Try it

Apply the risk triage to ten AI image ideas for a brand. For each medium- or high-risk idea, write what review, consent or alternative approach would be needed.

- [Previous: Copyright and ownership of AI images](https://optimizeall.com/learn/ai-image-generation-and-design/copyright-and-ownership)
- [Next: Labeling, disclosure and provenance](https://optimizeall.com/learn/ai-image-generation-and-design/labelling-and-disclosure)
- [All lessons of AI Image Generation and Design](https://optimizeall.com/learn/ai-image-generation-and-design)