EU AI Act mandates labeling for architectural visualizations
Realistic AI renderings now require 'deepfake' disclosure
1New regulation takes effect
Article 50 of the EU AI Act came into force on August 2. The provision mandates labeling for realistic-looking AI-generated images, audio, or videos to prevent deception-based harms. According to Szymon Sieniewicz from Addleshaw Goddard, architectural visualizations of buildings and interiors fall under the 'deepfake' category. Violations could lead to fines up to €15 million. The rules aim to maintain transparency in AI-generated content across industries.
2Why architectural renders are included
While deepfakes typically refer to synthetic media depicting real people doing things they never did, the law adopts a broader definition. Any realistic depiction of existing objects, places, or entities that could falsely appear authentic is covered. This includes AI renders of unbuilt buildings. Highly realistic city environments in video games will similarly be affected. The law focuses on preventing potential deception through synthetic media.
3Labeling requirements explained
The law requires clear disclosure of AI generation or manipulation at the point of first interaction. The European Commission has developed public-use icons for this purpose. Addleshaw Goddard advises using phrases like 'AI-powered tool' for maximum compliance. Basic edits like cropping or color correction don't require labeling. For text content, only public interest matters need AI disclosure, excluding general web copy.
4Product imagery impacts
Designers and companies must label significant AI alterations to product photos, such as making items appear more attractive, higher quality, or different colors. Simple retouching like background replacement remains exempt. For text, only content informing public interest matters requires disclosure unless substantially human-reviewed. Customer service chatbots must also indicate their AI nature upfront under Section 50(1).
5Artistic works get exemptions
The Act provides lighter disclosure rules for artistic, creative, satirical, or fictional works. However, Addleshaw Goddard notes that architectural visualizations and advertisements with promotional purposes won't qualify. Sieniewicz states, 'Standard architecture and design businesses will need to label AI-generated content.' The distinction aims to balance creative freedom with consumer protection against deceptive practices.
6Obligations for AI tool makers
Other provisions require tech companies like ChatGPT developers to embed machine-readable AI markers in synthetic content, which users cannot remove. This traceability measure helps identify AI-generated material across platforms. Service chatbots must clearly communicate their AI nature during initial interactions. These requirements reflect growing regulatory focus on AI transparency throughout the content lifecycle.
7Industry debates continue
The implementation may be welcomed by architectural visualization artists who argue AI diminishes their craft and produces error-filled results. Controversies persist about AI systems training on human-created works without permission. Concerns also exist that generative AI output might lack copyright protection, enabling unrestricted copying. These issues highlight tensions between innovation and creative rights in the AI era.
8Compliance timeline expectations
Sieniewicz doesn't expect immediate severe fines despite potential penalties up to 3% of annual turnover. Businesses likely will receive reasonable adaptation periods. Architecture firms and design studios should review AI usage policies. The European Commission's Implementation Guide will provide sector-specific clarifications. This phased approach aims to facilitate compliance while maintaining the Act's consumer protection goals.