AI Act Transparency Documentation

1. Risk Classification

Katoads is classified as a limited-risk AI system under the EU AI Act (Regulation (EU) 2024/1689). The platform uses generative AI to produce advertising images, videos, and marketing copy. This does not constitute a high-risk AI system as defined in Annex III, as it does not involve biometric identification, critical infrastructure, education, employment, law enforcement, or other high-risk use cases. However, as a generator of synthetic content, Katoads is subject to transparency obligations under Article 50.

2. System Description

Katoads is a SaaS platform that leverages multiple AI systems to automate advertising creative production. The AI pipeline operates as follows:

• Product Analysis: Google Gemini analyzes product information (images, descriptions, URLs) to extract marketing-relevant features, target audiences, and key selling points.
• Prompt Generation: Google Gemini generates optimized image prompts incorporating psychological marketing leverages, cultural context, and composition instructions.
• Image Generation: Google Gemini (native image generation) produces advertising images based on the generated prompts, with the product as the focal point.
• Video Generation: fal.ai (Kling 2.6 Pro) generates short-form video ads from static images for social media platforms.

3. AI Models Used

4. Intended Use

The AI systems are deployed exclusively for the following commercial purposes:

• Generating advertising creative content (images, videos) for e-commerce products across social media platforms (TikTok, Instagram, Facebook, YouTube, Pinterest, LinkedIn, Google Ads)
• Analyzing product information to extract marketing insights and generate persuasive copy in 6 languages
• Applying brand identity overlays (logos, colors, typography) to AI-generated visuals

5. Known Limitations

• AI-generated images may occasionally produce visual artifacts, incorrect product representations, or unintended elements despite safety filters.
• Marketing copy generated by AI may not always be culturally optimal despite multilingual copy guides and language-specific quality filters.
• Generated content should be reviewed by a human before publication to ensure compliance with advertising regulations in the target market.
• The AI does not have real-time knowledge of current advertising regulations, which vary by jurisdiction and platform.

6. Transparency Obligations (Article 50)

In compliance with Article 50(2) of the EU AI Act, Katoads implements the following transparency measures:

• All outputs generated by AI are clearly identifiable as AI-generated within the platform interface. Generation metadata records the AI models used.
• Users are informed at the point of generation that the content is produced by artificial intelligence systems.
• Machine-readable metadata is embedded where technically feasible to allow downstream identification of AI-generated content.

7. Human Oversight

Katoads implements human-in-the-loop controls at multiple stages:

• Users review and approve all generated content before download or export — no content is published automatically to ad platforms.
• The interactive layout editor allows users to modify, adjust, or override any AI-generated element (text, positioning, styling) before finalizing.
• Quality filters (forbidden element detection, text deduplication, language-specific copy validation) automatically flag potential issues for user review.

8. Risk Mitigation Measures

• Content safety: Gemini's built-in content safety filters prevent generation of harmful, illegal, or inappropriate advertising content.
• Anti-deception: Forbidden-elements system rejects visual clichés and misleading elements. Copy quality filters catch generic or misleading marketing phrases.
• Bias monitoring: Multilingual copy guides ensure culturally appropriate messaging and avoid stereotypical representations across 6 languages.
• Audit trail: All generation requests, inputs, outputs, and model versions are logged for traceability and compliance purposes.

9. Contact for AI-Related Inquiries

For questions about our AI systems, their classification, or this transparency documentation, please contact: contact@katoads.com