EU AI Act Code of Practice: How to Get Free Presumption of Conformity Before July 27

TL;DR - You have until July 27, 18:00 CEST to sign the EU Code of Practice on Transparency of AI-Generated Content. It costs 0 euro. It gives you a presumption of conformity with Article 50. Most founders do not know it exists.

Article 50 of the EU AI Act becomes enforceable on August 2, 2026. That is two weeks from today. If your SaaS generates content with AI, has a chatbot, or serves EU users, you need a compliance strategy.

The cheapest strategy costs zero euro and takes 15 minutes: sign the Code of Practice.

Here is what it is, who it's for, and how to sign before the July 27 cutoff.

Wait, there are TWO codes of practice?

Yes, and this is where most articles get it wrong.

The first Code of Practice (published July 2025) is for providers of general-purpose AI models -- companies like OpenAI, Anthropic, and Google that build the underlying models. That code covers safety, copyright, and transparency under Article 53 and 55.

The second Code of Practice (published June 2026, the one this article is about) is for providers and deployers of generative AI systems -- meaning you and me. It covers transparency of AI-generated content under Article 50.

Most coverage talks about the GPAI code. The Article 50 code is the one that affects micro-SaaS founders directly, and barely anyone is writing about it.

What does signing the Code actually give you?

A presumption of conformity with Article 50.

In plain language: instead of having to build your own argument that your AI transparency approach satisfies the law, you point to the Code. The burden flips. A regulator starts from the assumption that you are compliant, because you followed the EU's own blueprint.

Without that presumption, you are on your own. You need to document your marking and labelling approach, justify why it meets the legal standard, and hope a regulator agrees.

With the Code, you follow the playbook and you are done.

The European Commission and AI Board both assessed the Code and confirmed it is adequate (July 8-9, 2026). This is not a draft or a proposal. This is a live, endorsed compliance path.

Who can sign?

The Code is open to two groups:

Section 1 - Providers of generative AI systems (Article 50(2))

You build or deploy AI systems that generate synthetic audio, image, video, or text. This includes chatbots, image generators, content-writing tools, and any SaaS feature that produces AI output. You sign up for commitments on machine-readable marking and detection mechanisms.

Section 2 - Deployers of generative AI systems (Article 50(4))

You use generative AI in your operations and publish the output -- deepfakes, synthetic media, AI-generated text about public-interest matters. You sign up for disclosure and labelling commitments.

If you do both (build AI features AND publish AI-generated content), you sign both sections.

Technology providers of watermarking, provenance, and detection solutions can also sign Section 1.

What does the Code actually ask for?

The Code is structured around practical measures. Not theoretical principles.

For providers (Section 1), the commitments include:

• Machine-readable marking of AI-generated content (metadata, watermarking, or equivalent)
• Making detection tools available so downstream users can verify content provenance
• Documentation of the marking approach you implement

For deployers (Section 2), the commitments include:

• Visible labelling of deepfakes and AI-generated content about people, places, or events
• Disclosure when AI-generated text is published with the purpose of informing the public
• Maintaining records of compliance measures

These are things you need to do anyway starting August 2. The Code just tells you exactly what "good enough" looks like and gives you a legal presumption for following it.

Step-by-step: How to sign before July 27

Step 1: Download the signature form

Go to the EU digital strategy site and download the DOCX form: digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content

The form is also linked from the dedicated FAQ page.

Step 2: Fill it out

The form asks for:

• Your organisation's name and contact details
• Which sections you are signing (Section 1, Section 2, or both)
• A declaration that your organisation will implement the Code's commitments

It should be signed by a senior executive with authority to bind the company. For a micro-SaaS, that is you.

Step 3: Email it

Send the completed form to: CNECT-AIOFFICE-CODE-OF-PRACTICE-TRANSPARENCY@ec.europa.eu

Deadline: July 27, 2026, 18:00 CEST to be included in the initial published list of signatories.

Step 4: Actually implement the measures

Signing is a commitment, not a checkbox. You need to actually mark AI-generated content and label deepfakes in line with the Code's requirements. The initial list goes public before August 2, so a regulator or enterprise buyer can check who signed.

What happens after you sign

The EU publishes the list of initial signatories before August 2. Your company name goes on a public, timestamped list showing you committed to the Commission's own transparency standard before enforcement even started.

Enterprise buyers will see that list. Law firms advising procurement teams will see that list. It is a public signal that costs nothing and signals compliance maturity.

What happens if you do NOT sign

Signing is voluntary. Not signing is not a violation.

But you lose the presumption of conformity. You need to demonstrate compliance through other means. Market surveillance authorities will likely request more information from non-signatories to verify their approach.

In practice: the same obligation, more paperwork, more legal uncertainty.

What to do after signing (this is where Compliance Copilot comes in)

Signing the Code gets you the presumption. Actually staying compliant day-to-day requires ongoing work:

• Tracking which AI systems are in scope
• Maintaining machine-readable marking on every AI output
• Documenting the human-review workflow for deepfake exceptions
• Generating audit-ready transparency reports
• Updating risk classifications as you ship new features

This is what Compliance Copilot handles. Our EU AI Act module covers the full Article 50 compliance cycle:

• Risk classification for each AI system
• Article 50 checklist with 15+ checkpoints
• Human-review tracking and documentation
• Automated transparency report generation
• Audit-ready PDF exports

$49/mo for the full module. Founders Beta at $9/mo for the first 50 customers.

The deadlines that matter

The July 27 deadline is the one that expires first. And signing costs you nothing except 15 minutes and a commitment to marking your AI content, which you need to do anyway.

TL;DR

1. There are two EU AI Act Codes of Practice. The one for Article 50 transparency applies to you.
2. Signing gives you presumption of conformity with Article 50. This is the cheapest compliance you will find.
3. The deadline to be on the initial signatory list is July 27, 18:00 CEST (the EU extended it from July 22).
4. Download the form, fill it, email it, implement the measures.
5. After signing, use a compliance tool to stay on track day-to-day.

Check your EU AI Act compliance status ->

This article is informational, not legal advice. Consult qualified counsel for compliance decisions specific to your business.