AI in imagery: where to use it, where to avoid it

AI-generated images and videos provide unprecedented speed and quantity. One person can create dozens of options in a few hours, something that used to take weeks and a whole team. However, that speed comes with real risks, and most of the risks lie in the public release stage, not in the internal creation stage. Understanding that boundary is what you need before deciding where to use AI in your image process.

Two fundamentally different areas

You should divide the image process into two areas: the internal area and the public release area. AI works well and has fewer risks in the internal area. The public area requires an additional layer of scrutiny.

Three risks to check before launch

This is a checklist to review before hitting publish.

• Right of publicity: If your AI-generated images or videos use the face, body shape, or voice of a real person without written permission, even if reconstructed by technology, you are in a legally risky area. This applies to both celebrities and ordinary individuals. In Vietnam, Article 32 of the Civil Code 2015 requires the consent of the person whose image is used, and for commercial purposes, compensation must be paid unless otherwise agreed; the face and voice may also be considered personal data under the Personal Data Protection Law 2025. The concept of right of publicity mentioned above is from the United States, for reference only.
• Obligation to label AI-generated content: YouTube, Meta, and TikTok have all required the disclosure of AI-generated content in many cases. The European AI Act has Article 50 that mandates transparency for AI content. The specific application timeline needs to be confirmed through official documentation as the roadmap has many stages. In Vietnam, the Artificial Intelligence Law (No. 134/2025/QH15, effective from 01/03/2026) requires that audio, images, and videos generated by AI and made public must have identifiable signs when the content could mislead viewers into thinking it is real; the scope and method of labeling should be aligned with the guiding documents. If you run ads or post content on these platforms without labeling when required, your account and campaigns may be affected.
• Authenticity of products: AI images can look better than real products. The colors may be sharper, the materials smoother, and the proportions more balanced. When customers receive the product and see a difference from the image, trust is lost, and that is a long-term brand damage far greater than the benefits of a few quick photos. Legally, the Advertising Law and the Consumer Protection Law 2023 of Vietnam prohibit advertising information that misleads about the quality and utility of goods.

Most common errors

• Using a person's face or voice recreated by AI in commercial content without permission.
• No labeling of AI content when the platform is running requests, violating policies and potentially affecting the advertising account.
• Using AI-generated product images that are too perfect compared to the real items can disappoint customers when they receive their orders and leave negative reviews.
• Ignore signals that a segment of consumers is wary of AI content in advertising, according to some surveys. This is a qualitative signal, but for certain customer segments, especially those sensitive to authenticity, it can affect brand perception.

Advertising must be truthful, not misleading, and have evidence. When images or voices are generated by AI, the advertiser is responsible for ensuring that customers do not misunderstand the person speaking, the product being sold, or the nature of the content.

A summary of the core principles of the FTC (Federal Trade Commission) on truthful advertising for reference.

This article is for reference. The scope, pricing, and specific commitments of Sinh Vũ are detailed in the proposal and signed contract.