Chatbots and assistants · Voice agents and AI receptionists · AI agents · AI‑generated images, audio and video · Apps built on Lovable, Bolt or Replit
EU AI Act Article 50 has applied since 2 August 2026. It covers AI that talks to people and AI that generates content, and generators already on the market must mark their output by 2 December 2026. Get an honest, plain-English read on your exposure.
// recent audit Censory, an AI redaction tool, rewrote its privacy documents within two days of my audit.
// Why this matters now
The EU AI Act isn't coming — it's here, and the fines are real.
Maximum fine for Article 50 transparency violations — or 3% of global annual turnover, whichever is higher (Art. 99(4)). For SMEs, start-ups and small mid-caps, the lower of the two applies (Art. 99(6) and 99(6a)).
Non-EU founders included. If your product has EU users, Article 50 applies to you — regardless of where your company is incorporated.
The marking deadline. AI systems that generate audio, images, video or text and were on the market before 2 August 2026 must comply with Article 50(2) by 2 December 2026 (Regulation (EU) 2026/1744). The Commission approved its Article 50 guidelines on 20 July 2026. Enforcement runs through national market surveillance authorities (Art. 74).
// The [Logium] difference
A scanner reads the code without the law. A legal memo reads the law without the code. I audit your live product the way a regulator would — in text and in code.
// Is this you?
Your chatbot talks to EU users but never clearly identifies itself as AI at the start of the conversation.
You're using an LLM provider (OpenAI, Anthropic, Google…) but no Data Processing Agreement has been signed.
You're a US or non-EU founder who assumed EU law doesn't apply — it does, the moment you have EU users.
You're not sure what Article 50 actually requires and you've been waiting for "more clarity" before acting.
// The process
Answer 8 plain-English questions about your product. Get an instant traffic-light read on your Article 50 exposure — no sign-up, no email.
free · instantI review your live product: system prompts, UX copy, disclosures and DPA setup, through Upwork, which holds your payment in escrow while I work.
from $200A plain-English report with every gap, the exact fix, and a Now / Next / Later priority order. No legal fog.
actionable · no fog// What are you shipping?
Article 50 of the EU AI Act sets transparency duties for AI that talks to people and AI that generates content. Pick what you ship; each card goes to the matching fixed-price service on Upwork.
Article 50(1)
People must be told they are interacting with an AI, at the latest at the first interaction, unless it is obvious.
Start with the $200 verdict
Article 50(1) + 50(2)
Two duties can meet on one call: the caller must know they are speaking to an AI, and synthetic audio must be marked as AI‑generated by the provider of the system that generates it. Which one is yours depends on what you built.
Article 50(1)
An agent that talks to people, in chat, by e‑mail or on the phone, must make clear that it is an AI.
From $200
Article 50(2)
If your system generates synthetic images, audio or video, the outputs must be marked in a machine-readable format and be detectable as AI‑generated.
From $200
Article 50
The AI features inside an app you built yourself, such as a support chat or generated content, fall under the same rules.
From $200
// Audit packages
These three packages are for chatbots and AI assistants. Other products have their own fixed-price listings, linked from the cards above.
// not sure yet? start here
One product, seven questions by email, no calls. Three working days after your answers, you get a written verdict on which Article 50 duties apply and what follows if they do.
// flash-check
3-day turnaround
// full-audit
7-day turnaround
// compliance-pack
14-day turnaround
Runs through Upwork — escrow-protected. Upwork holds your payment in escrow while I work. Fixed price, agreed before anything starts, so you know the number up front.
// Who does the audit
// andrei.brisc
I spent four years as in-house legal counsel for a public authority in the EU, on EU-funded projects and public procurement. That is where I learned to read a regulation the way the people enforcing it do: word by word, condition by condition.
Then I crossed to the other side and learned to build AI systems. Today I audit chatbots and AI products the way a regulator would: reading the disclosure wording and the system prompt, checking the UX flow and the DPA status. Both sides, same audit.
Every audit I deliver is written by me — not a template, not a junior sub-contractor. You get one person who is accountable for every finding.
// Proof
One real result, named with the client's permission, and one full sample verdict you can read end to end.
// real audit · named with permission
Censory, an AI redaction tool, rewrote its privacy documents within two days of my audit.
Visit censory.app →// sample verdict · Article 50(1), 50(2), 50(4)
Every duty checked against the stated facts, the legal text quoted word for word, and a closing section on what the company is not required to do.
Read the sample verdict →// Common questions
The self-check takes 60 seconds. It is free, requires no sign-up, and shows you where you stand. If you're exposed, you'll know what to fix — and in what order.
Run the free 60-second self-check