AI, edge cases, and the ethics of creativity

(Or: What happens when the muse has a motherboard.)

AI has officially entered the chat, and the pitch deck, and the creative brief, and, let’s be honest, the panic dreams of half the creative industry. Depending on who you ask, it’s either the future of innovation or the end of originality. For agencies, the tension runs deep: we’re built on ideas, but we’re also built on deadlines. AI promises both speed and brilliance—but at what cost?

So here’s the question we keep circling back to: when does AI cross from helpful tool into ethical gray area—or outright theft?

Let’s talk about the spots where things get fuzzy. Most of the ethical tension in AI isn’t found in obvious misuse—it’s in the edge cases that feel almost fine.

An illustrator opens an image generator and sees a familiar style staring back. It’s their own, lifted from years of work, scraped from the internet, and repurposed into prompts. Is that inspiration or exploitation?

If AI models are trained on unlicensed art, do they inherit the ethics of their data? Agencies have always carried much of the responsibility for ensuring that the artwork used for client projects won’t get the client sued. Clearances and rights management have long been part of the hidden back-end of creative work. AI has complicated those efforts, and there are few or no regulations around it.

Working with ethical vendors that publish their training practices clearly is important—but increasingly difficult, time-consuming, or impossible. Hence the reason current legal advice often includes disclosing AI use to clients and defining who’s responsible for approving generated image use.

A client asks for AI-enhanced content, meaning faster, cheaper, more, more, more. The result technically “works,” but it’s missing the spark that made it human—and often, on brand. Whose work is it now: the agency’s, the algorithm’s, or the client’s accountant’s?

Writers are discovering AI tools that can mimic voices—sometimes their own. Imagine watching your style, your rhythm, even your phrasing, recreated by something that never lived a single one of your sentences. Efficient? Sure. Ethical? Not so much.

At this point, most agency people we talk to are tired of AI-generated copy for uses beyond basic production content. There’s a reason agencies tend to employ more than one writer. And no matter what anyone—or AI itself—tells you, writing style and copy personality are built in at the model level.

We’ve worked on fine-tuning with custom training data since the dawn of publicly available generative AI and large language models, and even the best-trained systems tend to slide toward mundane and generic without constant tune-ups.

AI-generated campaigns are already showing the same stereotypes humans have spent decades trying to unlearn. Turns out, if the inputs are biased, the outputs will be too. Machines may not mean harm, but they can replicate it with impressive consistency.

Agencies live in the pressure cooker of fast turnarounds and shrinking budgets. AI sometimes looks like an escape hatch—a way to do more with less. It also looks like an existential threat that could drown us all in a flood of mediocre content until clients can’t tell the difference anymore—and the public stops caring because it might all be fake.

Because with all its capabilities, AI is still creating things with very little grounding in reality, experience, or emotion. People don’t like feeling tricked into feeling something.

Creativity isn’t a production problem. It’s a connection problem.

AI can be a brilliant partner when used intentionally—it can spark ideas, help visualize concepts, or free up time for deeper thinking. But when used without thought or critique, it does what it was trained to do: blend, average, repeat. And nothing remarkable ever came from the middle.

So we ask ourselves: what does “authentic creativity” mean when a machine can mimic it? Maybe it means remembering that creativity has never been about the output—it’s about the insight behind it.

The creative world is having loud conversations about IP, bias, disclosure, and, well, em dashes. But the quieter questions are the ones that matter most:

• Should clients know when AI was used in the work they’re paying for?
• If a campaign concept started with AI, does that change how we value it?
• How do we ensure that “human touch” isn’t just a marketing line but a genuine part of the process?

There’s no universal code of ethics for AI in creative work yet—just a patchwork of principles and personal choices. Which means every agency, every team, and every creative has to decide where their own lines are.

At cat&tonic, we believe AI should assist, not replace. It’s a powerful tool, but it needs to be handled with integrity and curiosity.

We use AI to boost productivity, clarify thinking, augment research, and unlock creative possibilities that might otherwise stay trapped in draft folders—or in our heads. But we don’t use it to copy, exploit, or pretend. We don’t feed it unlicensed art, we don’t hand it our clients’ secrets, and we don’t let it write the story of a brand that’s meant to be human.

Our rule is simple: AI can make things, but only humans can make things meaningful. That includes the work lives of the humans we work with.

Because our work isn’t about producing volumes of content—it’s about telling the truth of who a brand is, in a way that moves people. And that’s something no dataset can replicate.