For much of the discussion surrounding the EU AI Act, attention has focused on large technology companies, developers of powerful AI models, and organizations deploying high-risk artificial intelligence systems.

As of August 2, 2026, the conversation has become much broader.

New transparency requirements under Article 50 of the EU AI Act extend obligations to a much wider range of organizations and professionals using AI-generated or AI-manipulated content. Advertising agencies, media organizations, designers, content creators, freelancers, businesses, NGOs, and other organizations may now need to disclose when people are interacting with AI or when certain content has been generated or substantially manipulated using AI.

The important shift is this: AI compliance is increasingly becoming an operational issue, rather than something reserved for Big Tech or highly regulated AI applications.

And because the rules can also apply to organizations outside the EU when their AI systems or content are made available in the EU market, their practical impact reaches well beyond European companies.

What changed on August 2, 2026?

Article 50 introduces transparency requirements covering several common ways AI is already being used across businesses and digital products.

The obligations broadly address four areas:

1. AI systems interacting directly with people

When an AI system directly interacts with a person, providers must generally ensure that users understand they are interacting with AI.

This can apply to technologies such as:

customer-service chatbots;
AI voice assistants;
AI-powered hotlines;
AI avatars;
conversational agents;
AI companions;
certain robots and automated interfaces;
other AI agents designed to communicate directly with users.

For businesses introducing AI into customer support, sales, onboarding or internal services, transparency therefore needs to become part of product and UX design.

A chatbot that looks and behaves like a human representative, for example, should not leave users guessing whether another person is actually on the other side of the conversation.

2. AI-generated and manipulated content

Providers of AI systems capable of generating or manipulating synthetic image, audio, video or text content face requirements around making those outputs detectable in machine-readable formats.

This creates another layer of responsibility beyond what users can visually see.

AI provenance and detectability may increasingly need to be considered alongside the generation itself.

For companies building AI-enabled platforms, this means transparency cannot simply be added through a disclaimer at the end of the development process. It may need to be incorporated into the architecture and output mechanisms of the product.

3. Emotion recognition and biometric categorization

Additional transparency obligations apply to certain AI systems designed for emotion recognition or biometric categorization.

Where such systems are deployed, individuals exposed to them generally need to be informed.

This is particularly relevant for organizations exploring AI technologies involving cameras, behavioral analysis, employee experiences, physical environments or other systems capable of processing information about individuals.

4. Deepfakes and AI-generated public-interest content

The fourth major category concerns AI-generated or manipulated content, including deepfakes and certain AI-generated text concerning matters of public interest.

In these situations, deployers may need to clearly disclose that the content was artificially generated or manipulated.

The definition of public-interest information can cover a broad spectrum, including subjects related to:

public administration;
justice and law enforcement;
fundamental rights;
public security;
healthcare;
environmental protection;
consumer safety;
economic matters.

That makes Article 50 relevant far beyond traditional technology teams.

Marketing, communications, media, PR, design and content departments may all need to understand how the rules affect their everyday workflows.

What does this mean for creators and marketing teams?

Generative AI has become deeply integrated into creative workflows.

A designer may use AI to replace an object in an image. A marketing team might generate product imagery. A copywriter could substantially restructure an article using an LLM. A video editor could create a scene that never happened. A podcast producer might synthesize someone’s voice.

Under the new transparency framework, some of these uses may require disclosure.

The distinction is generally between AI being used as a supporting editing tool and AI substantially generating or altering the meaning or representation of content.

Minor changes such as spellchecking, grammar correction, cropping, sharpening, basic color correction, video stabilization or similar technical improvements will generally be treated differently from substantial AI manipulation.

Replacing someone’s face, materially altering someone’s appearance, inserting people or objects into a scene, generating realistic events that never occurred or synthesizing realistic speech in a specific person’s voice are much more likely to trigger transparency considerations.

For marketing teams, this makes one question increasingly important:

At what point does AI assistance become AI-generated or AI-manipulated content that requires disclosure?

The answer may depend heavily on how the technology is being used and the context in which the resulting content is published.

Deepfakes are an obvious case, but the definition matters

Deepfake requirements are particularly significant because synthetic media is becoming easier and cheaper to produce.

AI-generated content representing real people, executives, celebrities or politicians can fall within these requirements when it could reasonably appear authentic.

The same principle can extend to synthetic voice recordings and realistic representations of products, objects, locations or events.

For companies, this has implications beyond obvious misinformation scenarios.

Imagine an advertising campaign using an AI-generated product image that significantly improves how the real product appears. Or a campaign featuring a synthetic representation of a person who appears to have participated in something that never occurred.

These are increasingly compliance questions as much as creative ones.

Certain exemptions and modified requirements exist for artistic, creative, satirical and fictional works, but professional and commercial contexts still require careful assessment.

Personal AI use and professional AI use are different

The EU framework provides exemptions for natural persons using AI for purely personal, non-professional activities.

That distinction becomes less straightforward when personal and professional activity overlap.

A person editing an AI-generated image privately is in a different position from an influencer publishing AI-manipulated content as part of a commercial partnership.

Similarly, freelancers, consultants and creators whose personal profiles form part of their commercial activity need to consider whether something that looks like “personal content” is actually connected to professional or economic activity.

The same technology can therefore create different obligations depending on how and why it is being used.

Who is responsible when an agency or contractor uses AI?

This is one of the most relevant questions for companies outsourcing marketing, software development, design or content production.

The European Commission’s guidance recognizes that responsibility depends partly on who controls the use of the AI system.

When an advertising agency independently decides whether and how AI is used while delivering work to a customer, relevant transparency obligations may fall on the agency rather than automatically on the customer.

Within organizations, individual employees operating AI systems under the company’s authority are generally not treated as separate deployers. The legal entity remains responsible.

The same can apply when contractors or freelancers operate AI systems on behalf of an organization and under its responsibility and control.

For companies, this makes governance important.

Contracts, procurement procedures and supplier policies increasingly need to answer questions such as:

Who decides when AI is used? Who controls the system? Who verifies the output? Who is responsible for disclosure?

These questions should be answered before AI-generated material reaches customers or the public.

A simple “we use AI” notice may not be enough

Transparency needs to be meaningful to the person encountering the system or content.

According to the Commission’s guidance, approaches that may be insufficient when used alone include disclosures hidden inside terms and conditions, metadata invisible to users, ambiguous terminology such as “assistant”, generic statements that a website “uses AI”, or highly technical descriptions such as saying that a service uses LLMs.

Providers may also have technical responsibilities to make generated outputs detectable in machine-readable formats.

Deployers, meanwhile, may need to ensure that people can clearly recognize when relevant content has been AI-generated or manipulated.

In practice, transparency therefore has both a technical layer and a user-experience layer.

What this means for companies building or deploying AI

For CTOs and technology leaders, Article 50 reinforces an important reality: AI governance needs to move closer to product development.

Compliance cannot live exclusively in a legal document.

Organizations building or integrating AI should start mapping where AI touches customers, employees, content and business processes.

That includes examining:

AI chatbots and conversational interfaces;
customer-support agents;
AI-generated marketing content;
synthetic images, video and audio;
automated content-generation workflows;
emotion or biometric technologies;
AI tools operated by agencies and contractors;
internal systems whose outputs eventually reach customers.

The next step is determining which party acts as the provider or deployer in each workflow and what technical and disclosure mechanisms are required.

This becomes particularly important in complex software environments where several AI providers, APIs, internal applications and external partners may contribute to a single customer experience.

AI transparency should become part of system design

The broader lesson from Article 50 goes beyond adding labels to AI content.

As regulation matures, organizations will increasingly need to know where AI operates inside their technology stack, what it produces, who controls it and how those outputs reach people.

That requires traceability.

Companies building new AI capabilities should consider transparency requirements during architecture and product design rather than treating compliance as a final checklist before launch.

For existing systems, the first priority should be visibility: mapping AI usage across applications, workflows, suppliers and content pipelines.

At Control F5 Software, we see this as part of a wider shift in enterprise AI.

The challenge is moving from simply asking “Can we use AI here?” to asking better questions:

How should AI fit into this system? What happens when it produces an output? Who remains accountable? How can users understand when AI is involved? And can the organization trace what happened when something goes wrong?

Those questions sit at the intersection of software architecture, AI governance and operational design.

And as the EU AI Act moves further into implementation, they are becoming part of building production-ready AI in Europe.

From AI experimentation to accountable AI systems

The EU’s new transparency requirements show how quickly AI governance is moving into everyday digital operations.

The impact is no longer limited to companies developing frontier models or deploying systems traditionally classified as high-risk. Businesses using AI for customer interaction, content creation, marketing, automation and digital services increasingly need to understand their responsibilities too.

For technology leaders, the practical response is not to slow AI adoption.

It is to build AI systems with transparency, accountability and control already embedded in them.

Because the companies best prepared for the next stage of AI adoption will be those that can innovate quickly while still understanding exactly where AI sits inside their systems, what it is doing and who is responsible for the outcome.

Source