AI-Generated Content Has Entered Its Accountability Era
What Organisations Should Do Now
For the past two years, much of the conversation around generative AI has focused on speed: faster copy, faster images, faster video, faster ideas.
That conversation is changing.
Since 2 August 2026, important transparency provisions of the EU AI Act have started to apply. Among other requirements, certain AI-generated or manipulated content must be identifiable, deepfakes must be disclosed, and people interacting with certain AI systems such as chatbots must be informed that they are dealing with AI.
For communications and marketing teams, this is not a reason to stop using AI. It is a reason to become more deliberate about how it is used.
The organisations that benefit most from AI are unlikely to be those that simply generate the most content. They will be the ones that know when automation adds value, when human judgement is essential, and when authenticity is part of the message itself.
That distinction matters particularly in pharma, healthcare, education and large corporate environments, where credibility, accuracy and stakeholder trust carry more weight than content volume.
AI IS BECOMING PART OF THE PRODUCTION WORKFLOW
Generative AI is already useful well before anything reaches an audience.
A communications team can use it to organise interview research, explore campaign angles, summarise lengthy background material, generate first-pass transcripts, create rough story structures, produce caption drafts or identify potential cutdowns from a longer piece of content.
Video teams can use AI-assisted tools for transcription, audio cleanup, searching footage, reframing, captioning and other repetitive post-production tasks.
These uses are fundamentally different from asking a model to fabricate a spokesperson, generate a realistic employee testimonial or create an event that never happened.
The technology may sit under the same broad AI label, but the communication risk is very different.
That is why organisations now need something more useful than a blanket policy of either "use AI" or "do not use AI". They need to understand the role AI is playing in each piece of content.
TRANSPARENCY IS NOW A PRACTICAL CONSIDERATION
The European Commission states that Article 50 transparency obligations under the AI Act apply from 2 August 2026. The rules cover areas including interactions with AI systems, machine-readable marking of certain AI-generated or manipulated material, and disclosure of deepfakes and certain AI-generated public-interest content.
The Commission has also published a voluntary Code of Practice on Transparency of AI-generated Content to help organisations and AI providers apply these requirements.
Not every use of AI suddenly requires a large warning label. The rules are more specific than that, and organisations should take appropriate legal or compliance advice where necessary.
But there is a broader communication lesson that applies regardless of the legal threshold: audiences should not have to wonder whether something presented as real actually happened.
For a communications team, that is a useful principle to build around.
THINK IN TERMS OF ASSISTANCE VERSUS REPRESENTATION
One practical way to approach AI is to ask a simple question:
Is AI helping us produce the content, or is AI replacing the reality the content claims to represent?
Using AI to transcribe a real interview is assistance.
Using it to help turn that interview into several draft social captions is assistance.
Using AI-assisted noise reduction on audio recorded in a busy environment is assistance.
Generating a realistic employee who appears to describe working at your organisation is representation.
Creating synthetic footage of a laboratory and allowing viewers to assume it is your facility is representation.
Cloning an executive's voice for a message they never recorded moves into even more sensitive territory.
The closer AI gets to representing a real person, event, workplace or experience, the more carefully authenticity, consent, disclosure and governance should be considered.
THIS MATTERS EVEN MORE IN REGULATED ENVIRONMENTS
Life sciences organisations already operate within systems built around documentation, review, traceability and risk management.
The European Medicines Agency has also addressed AI across the medicinal-product lifecycle, noting that its use needs to be considered alongside existing medicines regulation, data protection and wider EU legal requirements.
That does not mean every AI-assisted communications workflow becomes a regulatory project. It does mean that the culture of "generate first, figure it out later" is a poor fit for organisations where information accuracy and provenance matter.
Communications teams can borrow a useful idea from regulated workflows: maintain a clear chain of responsibility.
Who supplied the source information? Was AI used to generate or merely assist? Who checked factual claims? Who approved the final piece? Does the audience need to know that synthetic media was used?
Those five questions can prevent a surprising number of problems.
REAL PEOPLE MAY ACTUALLY BECOME MORE VALUABLE
There is an interesting consequence to the rapid improvement of synthetic media.
As generated images, voices and presenters become easier to produce, genuine access becomes more distinctive.
A real engineer explaining a difficult project, a scientist talking about why their work matters, an employee describing a workplace initiative, or a student sharing an actual experience contains something generative AI cannot manufacture reliably: evidence of a real organisation and its people.
That does not mean every corporate video needs to become a talking-head interview. It means organisations should recognise authenticity as an asset rather than an inconvenience.
The imperfections of real environments can sometimes strengthen credibility. A genuine production floor, laboratory corridor, classroom, event or team discussion gives context that polished synthetic imagery may lack.
AI can then work around that authentic core: helping organise the material, produce versions, create captions, repurpose interviews and speed up post-production.
That is a much stronger combination than replacing the core itself.
BUILD A SIMPLE AI CONTENT POLICY BEFORE YOU NEED ONE
A useful internal policy does not need to begin as a 40-page document. Communications teams can start with a simple traffic-light approach.
Green: AI-assisted tasks with low audience risk, such as brainstorming, transcription, internal summaries, caption drafts, metadata and production administration.
Amber: audience-facing generative work requiring human review, such as AI-generated illustrations, translated voiceovers, significant image manipulation or generated sections of marketing copy.
Red: synthetic representations that could reasonably be mistaken for reality, particularly fake testimonials, cloned voices, fabricated documentary footage or content involving sensitive personal, medical or confidential information.
The exact categories will vary by organisation, but defining them before a deadline arrives makes decisions much easier.
It is also worth recording which AI tools are approved, what information employees are allowed to upload, and who is responsible for reviewing public-facing outputs.
THE NEXT COMPETITIVE ADVANTAGE IS NOT SIMPLY SPEED
AI will continue to make content production faster. That part is unlikely to reverse.
But when everyone has access to similar generation tools, speed itself stops being much of a differentiator.
Judgement becomes the differentiator.
Knowing what is worth communicating. Knowing which stories need a real person. Knowing when a synthetic element should be disclosed. Knowing when an AI-generated shortcut could undermine the trust the organisation has spent years building.
For communications teams, the opportunity is therefore bigger than producing more material for less effort. It is using AI to remove friction around the work while protecting the human evidence at the centre of the story.
For video in particular, that may prove to be the most useful role for AI: not replacing reality, but helping organisations make more from the real stories, people and expertise they already have.
PRACTICAL TAKEAWAYS
• Audit where generative AI is already entering your communications workflow.
• Separate low-risk AI assistance from synthetic content that represents people, places or events.
• Keep human review for factual and audience-facing material.
• Establish rules for confidential, personal and regulated information before staff upload anything to external AI systems.
• Decide how your organisation will disclose synthetic or significantly manipulated media where required or appropriate.
• Preserve original source footage and approved versions when provenance matters.
• Treat genuine access to your people and workplace as a communication asset, not something AI automatically needs to replace.
