The EU AI Act is the European Union’s law on artificial intelligence, adopted in 2024 and applied in stages over the following years. For marketing and PR teams, its most relevant parts are the bans on manipulative AI practices, the transparency duties for chatbots and synthetic media, and the AI literacy duty for staff. An agency running campaigns that reach EU audiences has to build these into briefs, production and approvals.
What does the EU AI Act cover?
The Act sorts AI systems by risk. Some practices are prohibited outright, such as systems that use subliminal or manipulative techniques to distort behaviour in harmful ways, or that exploit vulnerabilities linked to age or disability. High-risk systems, such as those used in hiring or credit scoring, face strict obligations.
Most marketing tools fall into the limited-risk or minimal-risk groups. Limited-risk systems carry transparency duties. General-purpose AI models, such as the large language models behind ChatGPT, Claude and Gemini, have their own obligations, which fall mainly on the model providers rather than on agencies using them.
The Act applies outside the EU too. A company in Istanbul that runs an AI chatbot for German customers, or publishes AI-generated video to French audiences, falls within its reach for those activities. Check current guidance from the European Commission’s AI Office, since application dates and guidelines are still being published.
Which rules matter most for marketing and PR?
- Chatbots and AI assistants: people must be told they are interacting with an AI unless it is obvious.
- Deepfakes: image, audio or video content that resembles real people, places or events and could appear authentic must be disclosed as artificially generated or manipulated.
- AI-generated text published to inform the public on matters of public interest must be disclosed, unless it has been through human editorial review and someone holds editorial responsibility.
- Machine-readable marking: providers of generative systems must mark outputs so they can be detected.
- AI literacy: organisations using AI must take steps to ensure staff have sufficient understanding of it.
- Prohibited practices: manipulative or exploitative techniques are banned regardless of the tool.
The deepfake and chatbot rules are the ones that show up most often in day-to-day campaign work. The literacy duty is the one most often forgotten.
How Alien Road applies the EU AI Act in client campaigns
Our PR and campaign teams treat the Act as part of the brief, not a final legal check. Each step below runs on every campaign that reaches EU audiences.
Discovery: mapping AI use
At kickoff we list every place AI will touch the campaign: copy drafting, image or video generation, voiceover, chatbots, targeting tools and translation. Each item is tagged with the markets it will reach. This takes an hour and prevents surprises at launch.
Production: disclosure built in
Any synthetic image, video or voice that could be mistaken for real gets a disclosure plan before production starts: on-screen text, caption wording, and the platform’s own AI label. Real people are never generated without written consent. Chatbots open with a clear statement that the user is talking to an AI assistant.
Approval: two separate checks
Final assets go through a brand check and a compliance check. The compliance check confirms disclosure is present, no prohibited techniques are used, such as fake scarcity aimed at vulnerable groups, and the client’s legal team has seen anything sensitive.
Reporting and records
We keep a record of which AI tools were used for each asset, the prompts for generated media and who approved it. If a regulator, platform or journalist asks, the client can answer quickly.
What does this mean for PR specifically?
PR teams publish text meant to inform the public, such as press releases, expert commentary and thought leadership articles. The Act’s text disclosure rule has an exception for content that has gone through human review with editorial responsibility. That makes a named human editor on every release more than good practice; it is part of the compliance picture.
Synthetic spokespeople and AI-generated quotes are a clear risk. A quote attributed to a real executive must come from that person. An AI avatar presenting company news must be disclosed as such.
How do you meet the AI literacy duty?
The Act does not prescribe a course. It asks organisations to take measures appropriate to their staff and context. For a marketing team, that can be a short internal training covering how the tools work, their typical failure modes, data rules and disclosure duties.
- Run a session when someone joins and a refresher each year.
- Use real examples from your own campaigns.
- Keep attendance records.
- Update the material when tools or rules change.
- Include freelancers and contractors who produce content for you.
What are common misreadings of the Act?
Some teams assume the Act bans AI-generated ads. It does not. Others think it only matters for tech companies, when the transparency rules apply to anyone deploying chatbots or publishing deepfakes. A third misreading is that a platform’s automatic watermark covers the brand’s duty; watermarks get stripped, and the deployer still has to make sure people are informed.
The Act also sits alongside other laws. GDPR still governs personal data in targeting and personalization, and consumer protection law still governs misleading claims. Compliance with the AI Act does not settle those questions.
How should you prepare for later deadlines?
Obligations are being phased in, and the Commission has issued guidelines and codes of practice that fill in details. Assign someone to follow updates from the AI Office and your national regulator. A short quarterly note to the team on what has changed is usually enough for a marketing department.
Agencies and freelancers working for you are part of this. Ask them how they use AI on your account, and add the same disclosure and review steps to their contracts or briefs, so the rules hold across everyone who produces content in your name.
Where to start this week
List every AI tool your team used on campaigns for EU audiences in the last three months, with the market and asset type. Mark any synthetic images, video or voice that could pass as real, and any chatbot on your sites. For each, write down how the audience was told. Where the answer is “they were not”, add a disclosure step to the production checklist before the next campaign goes out.
Keep reading: PR Services · Digital PR