So everybody is happy to use AI to write or correct code for websites. And Bricks and Bricks are offering connectors to AI systems.
The European AI act wants AI generated content / code to be labeled, starting from 2026/08/02.
I suspect a lot of “shady” law companies will create systems to find AI generated content. Also i suspect a wave of cease-and-desist letters – especially in Germany, where i’m living.
How do / will you handle this topic?
Simple, don’t use AI. Learn to code. =]
AI generated content aswell.
From my understanding, there is no real legal obligation to do this unless we are creating an interactive AI experience like a chatbot or are publishing content meant to inform the public (like news or official information). In the case of the latter, having a human in the loop reviewing the page and content before publishing is key.
In either case, the upstream model connected to the MCP (Claude, Gemini, etc.) is the thing writing the code, not Bricks.
There is also a handy compliance checker here: EU AI Act Compliance Checker | EU Artificial Intelligence Act
This law is new to me, so I did a little bit of research. Public disclosure: I used AI to do the research.
Here’s what it had to say in the context of the new law and your comments:
The forum poster’s panic stems from a misinterpretation of Article 50 of the EU AI Act. While August 2, 2026, is indeed the enforcement date for the Act’s transparency rules, there is no legal requirement to label AI-assisted code or standard website copy
Key Misconceptions in the Forum Post
1. “AI-generated code must be labeled.”
- The Legal Fact: Article 50 of the EU AI Act contains an explicit exemption for AI tools performing an “assistive function for standard editing” or code completion. Using AI to write, correct, or refactor HTML, CSS, JavaScript, or PHP code—whether directly or via Bricks AI connectors—does not trigger any public disclosure or labeling obligation. Source code is not categorized as public synthetic media.
2. “All AI-generated text on a website must be labeled.”
The Legal Fact: Article 50(4) mandates text disclosures only for AI-generated text published with the specific intent to “inform the public on matters of public interest” (e.g., news reports, political opinion, or public policy).
Furthermore, the Act explicitly grants a Human Editorial Review Exemption: if AI-written text undergoes human editing, review, or oversight before publication, no disclosure label is required. Standard marketing content created for clients and edited by a human copywriter or developer is completely exempt.
3. “Shady law firms will send cease-and-desist letters (Abmahnungen).”
- The Legal Fact: Predatory German warning letters (Abmahnungen) rely on scannable, objective legal breaches under unfair competition law (UWG)—such as missing Impressum details or non-compliant Google Fonts embedding. Because using AI for code or edited web copy breaks no provision of the EU AI Act, there is no scrapable data or legal basis for warning letters.
What Actually Requires Labeling Starting August 2, 2026?
Under Article 50, website developers only need to implement transparency disclosures in specific, public-facing scenarios:
Interactive AI Tools (Article 50(1)): Public-facing chatbots, virtual support agents, or automated conversational widgets on a site must clearly inform visitors that they are interacting with an AI.
Deepfakes & Photorealistic Media (Article 50(4)): Artificially generated or heavily manipulated audio, video, or realistic images depicting real people or events must feature a visible label indicating synthetic media.
Upstream Technical Watermarking (Article 50(2)): This obligation falls on AI system providers (e.g., OpenAI, Midjourney) to embed machine-readable metadata into generated media, not on web developers using those tools.
Practical Checklist for Web Developers & Agencies
Maintain Human-in-the-Loop Workflows: Standard agency processes where human developers review code and human copywriters review text prior to site deployment fully satisfy the editorial exemption.
Label Front-End Chatbots: Ensure any client site deploying a conversational bot includes a clear label (e.g., “Powered by AI Assistant”).
Preserve Media Metadata: Avoid stripping technical metadata tags embedded in AI-generated images or media provided by AI tool suppliers.
What’s the functional difference between asking an LLM for a JS snippet versus copying one off Stack Overflow or a GitHub repo?
If code is generated dynamically at runtime, tracking AI usage matters for safety and auditing. But for static files authored during development, the AI is just a scaffolding tool. Once you review, refactor, and commit it, human accountability takes over.
Even low-code AI builders rely heavily on existing UI patterns and utility frameworks like Tailwind.