Have you had a moment recently where you realised just how easy this has all become?
I’m talking about writing paragraphs or emails in seconds, images conjured from a sentence, and campaign concepts sketched out before your cup of tea has even cooled.
It’s impressive, useful, and, at times, genuinely exciting. But quite often, a second thought follows closely behind: Where did this actually come from?
Not in the technical sense. We understand, broadly, how generative AI works, but in the human, creative, and ethical sense.
The truth is, AI-generated content doesn’t appear out of thin air. It’s built on something, on someone. Increasingly, we’re being asked to decide whether that matters, or whether we’re comfortable not knowing.
Why Provenance Suddenly Matters
In simpler times (or at least, simpler digital times), tracing content back to its source wasn’t especially difficult. For example: photos had photographers, articles had bylines, and designs had creators.
Even when things were shared, reposted, or remixed, there was usually a thread you could follow back to the beginning. People shared their sources. As a digital marketer, I was always taught to hyperlink any data I used.
AI complicates that thread or, more accurately, eliminates it.
Digital provenance, the ability to trace where something came from, how it was made, and who was involved, is becoming harder to establish just as it becomes more important.
There are efforts to rebuild that thread; these include watermarking systems, content authenticity standards, and initiatives like C2PA (Coalition for Content Provenance and Authenticity), which aim to embed provenance into the very fabric of digital media.
At the moment, it feels like piecemeal protection after the fact. It’s optional, inconsistent, and easy to remove if someone is determined enough.
This raises a slightly uncomfortable question. If we can’t reliably trace the origin of what we’re creating, sharing, or consuming…what does that do to trust?
Training Data is the Ethical Fault Line
Most of the current debate around AI ethics isn’t about what’s created. It’s where it comes from.
Large language models (LLMs) and generative systems are trained on vast datasets, including books, articles, images, and code, drawn from the entire internet. That data may be public domain, or it may be licensed, but in some cases, you can’t tell either way.
This is where things start to feel less abstract, because while AI companies often describe this process as “learning”, many creators and companies see it very differently.
To them, it’s extraction or appropriation. They see it as their work being absorbed into systems they didn’t consent to, can’t control, and aren’t compensated for.
That tension is no longer just philosophical; it’s turned legal.
When the Court System Gets Involved
Several high-profile cases are testing the boundaries.
The lawsuit brought by Sarah Silverman and other authors against OpenAI and Meta is one of the most widely discussed. It is a relatively simple claim that their books were used without permission to train AI models, potentially using pirated assets.
Then there’s The New York Times’ case against OpenAI and Microsoft, which shifts the focus from training data to AI-generated assets. The concern here wasn’t just that journalism was used in training, but that AI systems may reproduce or closely mimic that work in ways that undermine the original.
On the visual side, Getty Images filed suit against Stability AI. In this case, a similar question was asked: If a model has been trained on copyrighted images, including ones still bearing watermarks, where does inspiration end and infringement begin?
Even the entertainment industry is stepping in, with Disney, Universal (and later Warner Bros.) filing lawsuits against Midjourney for featuring copyrighted characters in its AI-generated images and videos. In this instance, they call it outright plagiarism.
Across all of these cases, the legal arguments differ in their detail, but the underlying issue remains consistent:
Can you build something new from someone else’s work without their knowledge, consent or compensation and still call it fair?
Until the courts or the legal system set a precedent, the lawsuits will continue. For now, the courts are still working that out, and they may be for some time.
Authorship, Reconsidered
Alongside copyright, there’s an equally significant shift happening around authorship.
Traditionally, authorship has been relatively clear. A piece of work belongs to the person who created it. With AI, that clarity starts to blur.
Is the author the person who wrote the prompt, the company that built the model, or the countless individuals whose work shaped the system in the first place? There isn’t a clean and clear answer, and perhaps that’s the point.
AI doesn’t create in isolation. It creates from the collective body of knowledge, culture, and creativity; a compressed version of the internet’s memory. That makes every output both original and derivative at the same time.
While that isn’t necessarily a problem, it challenges the way we assign value and credit to creative work.
What We Have is a Trust Problem
If provenance is about origin, then trust is about impact. Right now, trust feels fragile.
When we can’t easily tell whether something is AI-generated, human-made, or a blend of both, the default response isn’t always curiosity. It’s scepticism.
We’re already seeing it in small ways with the rise of “AI;DR”, a signal that people are switching off when something feels machine made.
For brands, that matters. For journalists, it matters even more. For anyone producing content professionally, it’s becoming unavoidable.
The risk isn’t just misinformation or deepfakes, though those are very real concerns. It’s something less obvious – the gradual erosion of confidence in what we’re seeing, reading, and sharing.
If everything might be synthetic, then nothing feels entirely certain. Once that uncertainty sets in, it’s difficult to reverse.
What Ethical AI Could Look Like
We’re well past the point of “just don’t use AI.” That ship has sailed. Instead, it’s how we use it in a way that feels considered, responsible, and defensible.
A few principles seem to be emerging, not as fixed rules, but as useful starting points on how we could use AI ethically in our businesses.
Transparency – Being open about when and how you use AI, and not as a disclaimer buried in the small print.
Consent – Giving creators more control over whether their work is included in training datasets in the first place.
Compensation – Explore (and using) models where contributors to AI systems are recognised and paid for their work.
Accountability – Clarify who’s responsible when something goes wrong, because “AI did it” isn’t going to fly.
Traceability – Invest in systems that make it easier to trace content back to its source.
I’m not saying any of these will be easy to implement. Collectively, they point towards a version of AI that works with the creative ecosystem rather than fraudulently extracting from it.
Where Do Brands Fit in All This?
For brands, creatives and agencies, this isn’t theoretical; it’s a practical issue. AI is already part of the workflow. Content is already being generated faster and at greater scale, often without visible inputs.
While this creates opportunity, it also creates responsibility. Your clients and audiences are already asking:
- Where did this come from?
- Can you prove it?
- Should you be telling us?
There’s a temptation to treat ethical considerations as something that can wait for regulations. Unfortunately, reputation tends to move faster than legislation.
Being thoughtful about how AI is used and able to explain it clearly will likely become a differentiator long before it becomes a requirement.
What happens next?
You won’t find a neat resolution with a set of rules that settle the debate all tied up in a bow. We simply aren’t there yet.
Global courts continue to work through details, while technology is quickly evolving, making the entire process even muddier. Standards are slowly emerging, yet without universal adoption.
What we have instead is a period of negotiation between:
- Innovation and responsibility
- Access and ownership
- What’s possible and what’s fair
While that negotiation plays out, the decisions are being made by companies, creators, and individuals. Those decisions will shape what this space looks like over the next few years.
A More Human Question
Despite the technical and legal complexity, this ultimately comes down to something quite simple: people make things.
They write, design, illustrate, photograph, and code, often for years, without much recognition, and always with care. AI isn’t going to replace that, but it definitely relies on it.
So, the question isn’t just whether we can generate content this way. It’s whether we can do it in a way that still respects the people and processes that made it possible in the first place (and we’re not talking about AI companies).
If we lose sight of that, if the origins of the ideas become too distant, too obscured, or too easy to ignore, then we risk losing something more fundamental than ownership.
We risk losing the connection between creativity and the people behind it, and that’s a thread worth holding onto.
