OPINION—By Demeke Kebede
Ethiopia Today, September 3,2026
You don’t have to narrow your eyes to see how artificial intelligence is changing things. Walk into a studio, a newsroom, a tech startups lab, a publishing house, it’s already there, doing work that used to take a team of people and a much bigger budget. That’s the appeal, and honestly, it’s a real one. Costs come down. Small players get their hands on tools that used to belong only to the big guys.
But there’s a question hiding under all that excitement that most people would rather not sit with: what happens when this machinery runs on somebody else’s work? Not abstract “data,” but a musician’s melody, a film maker’s footage, a journalist’s reporting, a photographer’s eye.
That’s the exact spot where artificial intelligence and digital piracy start to overlap, and it’s messier than most people realize.
A new generation of piracy/copyright infringement
Piracy has always kept pace with whatever technology comes along. First, it was burned CDs and DVDs passed hand to hand. Then file-sharing took over. Then illegal streaming and sketchy IPTV boxes. Each generation thought it had seen the worst of it. AI might prove them all wrong.
These systems now chew through unbelievable amounts of text, music, photos and film, and a lot of it is copyrighted. The fight happening right now, in courtrooms and parliaments and op-eds like this one, is about when a machine gets to “learn” from that material, what counts as real permission, and whether the people who made it in the first place ever see a cent.
Here’s the distinction worth holding onto: training an AI isn’t automatically piracy. But something being free to read doesn’t mean it’s free to profit from. A journalist doesn’t lose their copyright the moment an article goes live on a website. A musician doesn’t hand over ownership by putting a song on YouTube. The internet made copying easy. AI is what’s making copying, remixing and distributing something that happens at a scale nobody quite knows how to regulate yet.
From copying to cloning
And the threat doesn’t stop at reproduction. AI can clone a voice now. It can imitate an artist’s style closely enough to fool a casual listener. It can take a body of work and strip-mine it for parts, then hand you back something “new” built entirely out of someone else’s labor.
Sit with the economics of that for a second. A musician spends months, sometimes years, on an album. A journalist spends weeks chasing a story, sometimes at real personal risk. A filmmaker sinks years into a single project. If a machine can absorb all of that, learn its patterns, and spit out something that competes with it in seconds, with no transparency and no payment, what’s left to protect the people who actually did the work?
Copyright isn’t red tape standing between us and innovation. It’s the floor that creativity stands on. Take it away and you’re not left with more freedom, you’re left with fewer reasons to create anything in the first place.
Writing the rules of the game
None of this means we have to pick a side, AI or copyright. That’s a false choice, and a lazy one. What we actually need is an arrangement where both can exist without one eating the other.
Transparency has to come first. Tech companies shouldn’t get to build these systems in the dark, quietly, without ever saying what went into them. The EU has already moved on this: providers of general-purpose AI models are now required to keep a copyright policy on file and publish a real summary of what content trained their systems (Article 53, EU AI Act). And where protected work is being used commercially, there needs to be an actual functioning market for licensing it, not just goodwill.
There’s a strange kind of justice in the fact that the same technology causing this headache might end up being the fix. Watermarking, provenance tracking, automated detection, these tools are going to become the front line against digital infringement. And people deserve to know, plainly, whether what they’re reading came from a person or a machine. That’s exactly why the EU’s AI Act now mandates transparency and deepfake labelling, rules that take effect on 2 August 2026 (Article 50, EU AI Act).
None of this is a suggestion that Ethiopia should just copy what Europe is doing. The real lesson is simpler than that: AI policy and copyright policy can’t keep living in separate rooms.
Where Ethiopia stands today
Which brings this whole conversation home.
Ethiopia already has ground to stand on. The Copyright and Neighboring Rights Protection Proclamation (No. 410/2004), administered by the Ethiopian Intellectual Property Authority (EIPA), covers our literature, music, broadcasts, and software. And we’re not starting cold on the tech side either. In 2024, the Council of Ministers signed off on a National Artificial Intelligence Policy, and we passed our first real Personal Data Protection Proclamation No. 1321/2024.
Those are genuine steps forward, worth acknowledging. But there’s a gap between them that nobody’s closed yet. We don’t have a framework that says, plainly, who owns the creative content feeding into all of this, and under what terms.
And we’re not short on material. Our music, our manuscripts, our oral histories, our languages, and our film, it’s a genuine goldmine. Most of it is still barely digitized and completely disconnected from any modern rights system. Leave that door open and global AI systems will walk through it, harvest what they want, and leave nothing behind for the people who made it. That’s not a hypothetical. That’s just what happens when there’s no fence.
Building a digital rights infrastructure
At some point we have to stop treating copyright like something you only reach for after the damage is done, a lawsuit here, and a takedown notice there, and start treating it like actual infrastructure.
Picture something like this: a centralized, machine-readable registry run by EIPA. Every book, song or archival recording that gets digitized carries a permanent record attached to it, who owns it, what the terms are. Before any AI developer scrapes that content, their system has to check the registry first. Not a suggestion. A gate.
None of this means shutting AI out. If anything, it’s the opposite. Done right, AI could be one of the best things to happen to Ethiopian creators in a generation, translating our literature for readers who’d never otherwise find it, digitizing archives before they’re lost, and giving small creative businesses a shot at markets they couldn’t reach before.
But that only works if the creative economy underneath it survives. We can’t afford to spend the next decade as consumers of somebody else’s technology and unpaid suppliers of the raw material that makes it run. In this economy, content is data, and data is power. It’s time we built something that makes sure Ethiopian creators actually keep hold of theirs.
Editor’s Note: The opinion expressed in this article is that of the author and does not necessarily reflect the position of Ethiopia Today.