Back to Blog

AI Labelling Obligation from August 2026: What Changes and What to Do

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder at Superkind

Superkind logo on an abstract white waves texture

AI labelling obligation is the phrase keeping compliance officers awake since August 2026. Your chatbot answered a customer question, your marketing team published an AI-generated social media graphic, your sales assistant drafted a follow-up email - and none of it carried a disclosure. Under Article 50 of the EU AI Act (Regulation (EU) 2024/1689), all three need one. This article explains what the transparency obligations require, which deadlines apply, and what your team needs to do now - plus how Superkind’s ready-made AI team for compliance handles disclosures, content marking and audit trails automatically.

Key takeaways

Get a ready-made compliance team - Superkind comes with a compliance team (Disclosure Guard, Content Marker, Audit Logger) that labels, marks and logs every AI interaction automatically. Start for free with unlimited users, or build your own.

Keep control, keep data in Europe - hosted in Europe, GDPR compliant, aligned with the EU AI Act, no training on your data, you decide what needs approval. Sovereign and independent.

Four obligations since 2 August 2026 - chatbot disclosure, machine-readable content marks, deepfake labels and public-interest text labels1.

Fines up to EUR 15 million or 3 percent of turnover - whichever is higher. No SME exemption on the obligation itself7.

Grace period only for marking - providers of existing generative AI systems have until 2 December 2026 for machine-readable marks; everything else applies now2.

Your AI team for compliance

AI disclosure check
Disclosure GuardAI employee
Every outgoing AI message
Checked 14 outgoing AI messages this morning. 12 had correct disclosures. 2 customer replies were missing the AI label - added it before sending.
Messages checked14
Labels added2
Blocked for review0
Runs automatically on every AI-generated message
Example: Disclosure Guard at work in Outlook

What is the AI labelling obligation?

The AI labelling obligation is the requirement under Article 50 of the EU AI Act to make people aware that they are dealing with AI or AI-generated content. It applies to providers and deployers of AI systems that interact with people, generate synthetic content, perform emotion recognition or produce deepfakes1. The obligation took effect on 2 August 2026 and carries fines of up to EUR 15 million or 3 percent of worldwide annual turnover7. One example of a system built for this: Superkind, an AI employee that carries the disclosure in every interaction through Outlook, Teams, Slack and the web app, with an audit trail and 1,700+ integrations into tools like Zendesk, HubSpot and Notion, hosted in Europe.

The European Commission adopted guidelines on 20 July 2026 and published a voluntary Code of Practice for AI content marking34. The code is voluntary, the legal obligations are not. National market surveillance authorities enforce them, and the AI Office handles systems built on general-purpose AI models2.

How it works with Superkind: your AI team for compliance

You write to Superkind by email, in Teams, in Slack or in the web app, and Superkind does the task in your programs - with 1,700+ integrations, including Zendesk, Notion, HubSpot and SharePoint. It works with your tools instead of replacing them, hosted in Europe and GDPR compliant.

With Superkind you don't start from zero: it comes with ready-made AI employees for compliance. They run automatically in the background, and you decide where they ask you first. Use them right away, adapt them or build your own.

Disclosure Guard

AI employee for AI transparency disclosures

Every chatbot message, every AI-drafted email and every generated text needs a disclosure now. The Disclosure Guard checks every outgoing AI message for the right label before it goes to the customer. If a disclosure is missing, it adds it. Only edge cases with unclear rules land with you.

Works inOutlookSlackMicrosoft Teams+ 1,700 more
AI disclosure check
Disclosure GuardAI employee
Every outgoing AI message
Checked 14 outgoing AI messages this morning. 12 had correct disclosures. 2 customer replies were missing the AI label - added it before sending.
Messages checked14
Labels added2
Blocked for review0
Runs automatically on every AI-generated message
Example: Disclosure Guard at work in Outlook

Content Marker

AI employee for machine-readable AI content marking

Article 50(2) requires machine-readable marks on AI-generated images, audio, video and text. The Content Marker adds metadata tags to every piece of AI content before publication. It checks the format, logs the mark and files the proof in SharePoint.

Works inMicrosoft TeamsSharePoint+ 1,700 more
ContentMarketing
Content MarkerAI employee
New AI content in the pipeline
Marked 8 new blog images and 3 social media graphics with C2PA metadata. Proof logged in SharePoint.
Images marked8
Graphics marked3
Log updatedSharePoint
Runs automatically for every new AI-generated asset
Example: Content Marker at work in Microsoft Teams

Audit Logger

AI employee for AI transparency audit logs

When the regulator asks for proof, you need a trail. The Audit Logger records every AI interaction, every disclosure, every content mark and every approval in a structured log. Every Friday it sends you the weekly summary.

Works inSlackExcelSharePoint+ 1,700 more
#compliance
Audit LoggerAI employee
Routine - every Friday at 7:00
Weekly audit report ready. 312 AI interactions logged, all with disclosures. 47 AI-generated assets marked. 0 open items.
Interactions logged312
Assets marked47
Open items0
Runs automatically every Friday
Example: Audit Logger at work in Slack

Every employee is ready to use right away. Or build your own on the platform.

Safe by default

Superkind is built for the EU AI Act from day one: every AI interaction carries a clear disclosure, content is logged, and the audit trail sits in your systems. Sovereign and independent.

How Article 50 works

  1. Chatbot and AI agent disclosure (Art. 50(1)) - providers must design the system so that people know they are dealing with AI at the start of the first interaction, unless it is already obvious1.
  2. Machine-readable content marking (Art. 50(2)) - providers of generative AI must mark synthetic audio, images, video and text with machine-readable metadata (e.g. C2PA watermarks or embedded tags) so the content can be detected as AI-generated1.
  3. Emotion recognition and biometric categorisation (Art. 50(3)) - deployers must inform the people exposed to these systems. At the workplace and in education, emotion recognition is generally prohibited6.
  4. Deepfake and public-interest text labels (Art. 50(4)) - deployers must visibly label deepfakes and AI-generated text published on matters of public interest without substantive human editorial control1.

Deadlines

2 August 2026 - all Article 50 obligations apply. 2 December 2026 - grace period for machine-readable marking of generative AI systems already on the market before 2 August 20265. Content generated before 2 August 2026 does not need retroactive labelling, but content generated before that date and published on or after it does2.

Obligations at a glance

ObligationWhoWhat to doDeadline
Chatbot disclosureProviderTell users they talk to AI2 Aug 2026
Content markingProviderMachine-readable marks on synthetic content2 Aug 2026 (existing systems: 2 Dec 2026)
Emotion / biometricDeployerInform exposed people2 Aug 2026
Deepfake labelsDeployerVisible disclosure at first exposure2 Aug 2026
Public-interest textDeployerLabel AI text on public topics without editorial control2 Aug 2026

Source: Article 50, Regulation (EU) 2024/16891

Examples

Superkind: compliance built in - every interaction through Outlook, Teams, Slack or the web app carries the AI disclosure. Content generated by Superkind is logged with an audit trail. Approval flows let you review sensitive messages before they reach the customer. With 1,700+ integrations into Pipedrive, Freshdesk, Calendly and Excel, the proof sits where your team already works. Start for free at app.superkind.ai.

Customer service chatbot - a chatbot on your website answers questions about orders and returns. Under Article 50(1), the first message must tell the visitor they are talking to AI. A small banner or opening line is enough, as long as it is clear and comes before the conversation starts2.

AI-generated marketing images - your designer uses an AI tool to create product visuals for social media. Under Article 50(2), these images need machine-readable marks (such as C2PA metadata). If you publish them on matters of public interest, a visible label is also needed4.

AI-drafted email replies - your sales team uses an AI assistant to draft follow-up emails. If the AI interacts directly with the customer (sends the email on its own), disclosure is required under Article 50(1). If a person reviews and sends the email, the chatbot disclosure does not apply, but the content may still need marking under Article 50(2) depending on how it is generated5.

What it costs and what you need

Compliance itself does not have to be expensive. What you need:

  • An AI system inventory - list every AI system your team uses or provides, and map each one against the four obligation categories above5.
  • Disclosure procedures - for chatbots and AI agents, add a clear opening message. For content, set up machine-readable marking (the voluntary Code of Practice offers a route4).
  • An audit trail - record which AI generated what, when and for whom. Superkind does this automatically with every task.
  • AI literacy training - Article 4 of the EU AI Act requires you to support AI literacy for the people who use or are affected by AI systems1.

Data protection and EU hosting

Transparency obligations sit on top of your GDPR duties. Choose an AI system that keeps both simple: hosted in Europe, GDPR compliant, no training on your data, with a data processing agreement. Superkind offers all of this through its Trust Center. Sovereign and independent: hosted and processed in Europe under European law, AI models from OpenAI, Anthropic, Mistral and Google processed in Europe, no lock-in to one model vendor.

What Superkind costs

Free to start: a one-off up to 150 hours of work free to try, no credit card, unlimited users. Business from EUR 50 per month for 50 hours of work a month for the whole company with unlimited users. Enterprise with custom pricing, SLA and SSO. See the pricing page for details (as of November 2026).

How to start

  1. Inventory - list your AI systems and map them to the four obligations.
  2. Disclose - add the AI label to chatbots and AI agents. Add machine-readable marks to generated content.
  3. Log - set up an audit trail. Superkind does this from day one for every interaction in Outlook, Teams, Slack and the web app.
  4. Train - brief your team on what Article 50 means for their daily work (AI literacy, Article 41).
  5. Try Superkind - sign up for free at app.superkind.ai, give the first task in Teams, Slack or by email and see the disclosure and audit trail in action.

Conclusion

The AI labelling obligation under Article 50 is not a future concern - it applies now. Four categories, clear deadlines, real fines. The good news: most teams can meet the requirements with the right AI system and a short checklist. Start with the inventory, add disclosures and marks, set up an audit trail. Superkind handles all three from day one, hosted in Europe, with 1,700+ integrations into Freshdesk, Asana, Pipedrive and SharePoint. The fastest way to check: sign up for free and give Superkind a task.

Frequently asked questions

Article 50 of the EU AI Act (Regulation (EU) 2024/1689) requires providers and deployers of certain AI systems to make people aware they are dealing with AI. For chatbots, a clear disclosure at the start of the interaction. For AI-generated images, audio, video and text, machine-readable marks. For deepfakes and AI-generated public-interest text, visible labels. The obligations apply from 2 August 2026.

Four categories: (1) AI systems that interact with people, such as chatbots and AI agents, (2) generative AI that produces synthetic audio, images, video or text, (3) emotion recognition and biometric categorisation systems, and (4) deepfakes and AI-generated text on public-interest topics. If your company uses any of these, the transparency obligations apply to you as a deployer or provider.

Fines of up to EUR 15 million or 3 percent of worldwide annual turnover, whichever is higher. National market surveillance authorities enforce the rules. The proportionality principle may apply for smaller companies, but the obligation itself does not have an SME exemption.

Superkind is built for compliance from the start. Every AI interaction through Outlook, Teams, Slack or the web app carries a clear disclosure. AI-generated content is logged with an audit trail. With 1,700+ integrations into tools like Notion, HubSpot, Zendesk and SharePoint, Superkind keeps the transparency records where your team already works. Hosted in Europe, GDPR compliant, aligned with the EU AI Act.

Article 50(1) applies when an AI system interacts directly with a person, including employees. If your team uses an internal chatbot or AI assistant, the people using it must know it is AI. Internal drafts that are not published do not need the deepfake or public-interest text labels. But if AI-generated content is shared externally, the full obligations apply.

Only for the machine-readable marking obligation under Article 50(2): providers of generative AI systems already on the market before 2 August 2026 have until 2 December 2026 to add machine-readable marks. All other obligations - chatbot disclosure, deepfake labels, public-interest text labels - apply from 2 August 2026 with no grace period.

Related articles

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder of Superkind, the AI employee teams write to in Outlook, Teams and Slack. Henri helps companies, from small teams to the Mittelstand, hand their routine work to AI - without new tools, consultants or an in-house AI team.

Send Superkind your first compliance task.

Start for free in Outlook, Teams or Slack. Or let us show you in 30 minutes how it runs in your systems.