AI in media and journalism: transparency, deepfakes and media freedom

Short answer: AI affects media and journalism along three tracks you should not conflate. The AI Act (Article 50) requires transparency about artificially generated or manipulated content. The DSA regulates how platforms spread disinformation and illegal content. The European Media Freedom Act (EMFA) protects the independence of the newsroom and its sources. Together they shape how you use AI responsibly in news.

Transparency on deepfakes and AI text

Article 50 of the AI Act requires that deepfakes — realistic-looking but artificially created or manipulated image, audio or video — be labelled as such. Text published to inform the public on matters of public interest also carries a disclosure duty, unless the content has been human-reviewed and someone holds editorial responsibility.

For a newsroom this means: anyone using AI to generate imagery or draft text must make that visible to the reader. Human final review with clear responsibility is the exemption ground.

The DSA: distribution via platforms

The Digital Services Act targets not the creator but distribution. Very large online platforms must assess systemic risks, including disinformation and the impact on public information. They must be able to remove illegal content and be transparent about recommender algorithms.

For journalism this cuts both ways: platforms are distribution channel and risk source. AI-generated fake-news campaigns fall within the systemic risk that platforms must manage.

The European Media Freedom Act

The EMFA, applicable from August 2025, protects media pluralism and independence. Key points: protection of journalistic sources, safeguards against unjustified use of spyware against journalists, and transparency on media ownership.

AI touches this where editorial decisions are automated or where AI systems are used to track journalists. The EMFA sets the line: editorial autonomy must not become subordinate to technical or commercial steering.

What to do

• Label AI content: mark deepfakes and AI-generated text in line with Article 50.
• Secure human editing: record who holds editorial responsibility — that is the exemption ground for text.
• Check your distribution: know the DSA obligations of the platforms you publish on.
• Protect sources: ensure AI tools (e.g. transcription or cloud storage) do not undermine source confidentiality.
• Be transparent about ownership and use: this aligns with both the EMFA and the broader state of AI regulation.

In journalism, transparency is not an administrative burden but the core of credibility. See also AI in other sectors such as telecom, where the same transparency and security logic recurs.