EU AI Act Article 50 Takes Effect, Requiring Clear Labels for AI Content
The European Union has officially begun enforcing Article 50 of the EU AI Act, introducing new transparency requirements for companies that develop or use artificial intelligence systems across the region.
Under the rules, people must be clearly informed when they are interacting with an AI system. AI-generated or manipulated images, videos, audio, and text may also need to carry detectable labels identifying them as synthetic content.
The regulation is designed to address growing concerns about deepfakes, impersonation, fraud, consumer deception, biometric surveillance, and the large-scale manipulation of public opinion.
As generative AI becomes more realistic, distinguishing between human-created and machine-generated content has become increasingly difficult. Article 50 aims to give users greater awareness of when AI is involved.
What AI Providers Must Do
Companies that develop directly interactive AI systems must design them so users know they are communicating with a machine.
This disclosure may not be required when it is already obvious to a reasonably informed and observant person that the system is powered by AI.
Certain law enforcement systems may also be exempt when they are used to detect, prevent, investigate, or prosecute criminal offenses, provided appropriate safeguards are in place to protect the rights of third parties.
However, the exemption may not apply when members of the public directly interact with the system, such as when using an AI service to report a crime.
Providers of AI systems capable of generating or manipulating audio, images, video, or text must also ensure that their outputs contain machine-readable markings.
These markers should allow automated systems to detect whether the content was artificially generated or altered.
The EU requires these markings to be effective and interoperable whenever technically feasible. Regulators may consider implementation costs, available technology, and current industry standards when evaluating compliance.
Basic AI-assisted editing may fall outside the requirement when it leaves the user’s original material largely unchanged. Minor photo corrections, for example, may not require synthetic-content labels. Fully generated replacements or major AI alterations are more likely to be covered.
What AI Deployers Must Disclose
Businesses and organisations using emotion recognition or biometric categorisation systems must inform individuals who are exposed to the technology.
Any personal information processed by these systems remains subject to existing European data protection laws, including the General Data Protection Regulation.
Different rules may apply when the system is operated by an EU institution or used in a law enforcement environment.
Deepfakes Must Be Identified
Article 50 introduces specific disclosure requirements for deepfake content.
Images, videos, or audio that have been artificially generated or significantly manipulated must carry a clear notice explaining that AI was used.
Artistic, fictional, satirical, or creative works receive more flexible treatment. Disclosures are still required, but they may be presented in a way that does not unnecessarily disrupt the audience’s experience.
A movie containing AI-generated scenes, for example, may provide a suitable disclosure without placing a warning across every frame.
AI-Generated News and Public-Interest Content
The regulation also addresses AI-generated text published to inform the public about matters of public interest.
Publishers must disclose when this material has been generated or manipulated using AI.
However, disclosure may not be required when the content has undergone meaningful human review and a person or organisation accepts editorial responsibility for the final publication.
This means professional newsroom oversight may satisfy the rule.
Publishing raw, unverified AI-generated text directly as a public-interest news story may not.
The distinction highlights the continuing importance of human editors, fact-checkers, and accountable publishers in the age of generative AI.
Disclosures Must Be Immediate and Accessible
Required disclosures must appear no later than the user’s first interaction with or exposure to the AI system or content.
They must be clear, easy to distinguish, and accessible under existing European accessibility requirements.
Informing people only after they have already interacted with the system would not meet the standard.
Who Will Enforce the Rules?
Enforcement responsibilities will be divided among several European authorities.
National market surveillance authorities will handle most compliance cases.
The EU AI Office will supervise systems that fall under its jurisdiction, while the European Data Protection Supervisor will oversee cases involving EU institutions acting as AI providers or deployers.
The EU’s Preferred Compliance Approach
For AI-generated content labelling, organisations may follow the EU Code of Practice on Transparency of AI-Generated Content.
Companies that adopt the Code can use it as one method of demonstrating compliance with Article 50.
Participation is not mandatory. Organisations that choose not to follow the Code must show that their alternative labelling systems adequately meet the regulation’s requirements.
The final assessment will depend on the responsible enforcement authorities.
Other Article 50 duties, such as informing users that they are interacting with AI or disclosing the use of deepfakes, do not have the same compliance shortcut.
Companies must develop their own appropriate processes while using the European Commission’s guidance as a reference.
Providers and Deployers Have Different Responsibilities
A major part of compliance involves determining whether an organisation is acting as an AI provider, an AI deployer, or both.
A provider develops or supplies the AI system.
A deployer uses the system as part of its operations, products, services, or communications.
The responsibilities that apply will depend on the organisation’s role within the AI value chain.
A company that develops its own generative AI platform and also uses it to communicate with customers may carry obligations as both provider and deployer.
What This Means for Businesses
Companies operating in the European Union must now examine how their AI tools interact with users and how synthetic content is created, published, and distributed.
Key compliance areas include:
- Clearly informing users when they are communicating with AI
- Adding detectable markers to synthetic content
- Disclosing deepfakes and significant AI manipulation
- Informing people about emotion recognition or biometric categorisation
- Establishing human editorial review for AI-generated public-interest content
- Keeping records that demonstrate how transparency requirements are being met
The new rules are likely to affect AI developers, social media platforms, media organisations, advertisers, customer service providers, content creators, government agencies, and companies using generative AI in public-facing services.
Why It Matters to the Philippines
Although the EU AI Act applies primarily within the European market, its impact will extend far beyond Europe.
Filipino technology companies, business process outsourcing providers, AI developers, content studios, and digital service providers working with European clients may also need to comply.
The rules could also influence future AI policies in the Philippines, particularly in areas such as deepfake regulation, election-related content, AI-generated advertising, digital fraud, biometric technology, and media transparency.
Article 50 sends a clear message: using AI is not prohibited, but hiding its involvement from the public may no longer be acceptable.
As synthetic media becomes more convincing, transparency will be essential in protecting public trust, personal identity, and the credibility of digital information.
For businesses and creators, the era of unlabeled AI content is beginning to close.
