The AI copyright debate misses the bigger question - does the work actually perform?

Shelley Friesen, Founder of Melbourne Social Co and ATTN.CTRL

There have been a few moments lately where brands take the social-first content our team creates, plug it into an AI engine, and ask for ten more versions of it. It's not quite the same. It doesn't perform nearly as well. But it's given us reason to properly look at our own policies, where our IP is protectable, and where it isn't, because until recently, nobody had drawn that line clearly for any of us.

That's the part of this AI conversation I think gets skipped. Everyone jumps straight to whether it's right or wrong. I'm less interested in that argument than I am in the grey area itself, because grey areas are expensive. They cost you time in every client conversation, every internal debate about what's fair game and what isn't. A clear line is worth more to a business than a moral position.

This week, the federal government's new AI framework introduced protections around AI companies training on Australian books, music, art and news without consent. Headlines focused on data centres and power grids. I focused on the copyright line, because it's the first time I've seen this grey area get an actual edge instead of being left as a case-by-case judgment call every agency, brand and creator has been making up as they go.

Here's the thing that's actually interesting to me, more than the ethics of it: the AI versions still don't perform as well. Not close. Whatever a model produces when it's replicating a campaign, a tone, a visual style, it's faster, and it's cheaper, but the numbers don't lie, and the numbers keep saying the same thing. Original, human-made work outperforms the replicated version every time we've tested it side by side.

That's the argument I'd make to any brand thinking about this, not "you shouldn't," but "here's what the data shows you when you do." Speed isn't the same as performance. Ten versions of something that underperforms is still underperforming, just at volume.

I don't think this framework is about punishing anyone for using the tools available to them. It's about giving the whole industry a clearer starting point, so brands, agencies and platforms aren't each quietly figuring out where the edges are on their own. Less grey area is good for everyone. It means the conversations we have with clients about AI can be about what actually works, not about where the boundaries sit.

So here's where I've landed. Use the tools. Move fast where speed genuinely helps. But keep an eye on the results, not just the output. In every case we've tracked, the work made by people, not replicated by a model, is still the work that performs. That's not a policy position. That's just what the numbers keep telling us.

The government just made the boundary clearer. The data already made the case.