You will be able to explain why there is no single answer to who owns AI-generated content.
Kelvin handles marketing for his family's bakery in Joo Chiat. Last month he used an image generator to make a poster for the Mid-Autumn mooncake launch: a moon over shophouse rooftops, lanterns, a box of mooncakes in the foreground. It looked great. A friend who runs a print shop asked if she could use a similar style for her own posters. Kelvin realised he had no idea whether the image was his to object to, or anyone's at all.
He asked an assistant, and it gave him a confident two-line answer. That answer is the problem this lesson is about. Ask the same question in a different country, or about a different kind of work, or next year, and the honest answer can change.
Copyright law in most countries grew up around a simple picture. A person writes a book, takes a photo or composes a song, and the law gives that person rights over copies of it for a period of time. The whole system assumes there's a human author whose creative choices produced the work.
AI tools don't fit that picture neatly. When Kelvin typed a prompt and the tool produced the poster, who made the creative choices? He chose the subject and the mood. The tool decided the composition, the brushstrokes, the exact shade of the lanterns. The company that built the tool made choices too, and so did the thousands of artists whose work the model learned from.
Courts and regulators in several countries are working out how existing law applies to this. Some have given early views. The US Copyright Office, for example, has said it registers only material created by a human, and that applicants must disclose AI-generated content. Other countries are taking different approaches, some through courts and some through consultations on new rules. In Singapore, the Intellectual Property Office of Singapore, IPOS, is the place to check for current guidance.
Copyright is national. Each country has its own law, and international treaties only line them up in broad terms. So "who owns this?" can have a different answer in Singapore, the United States, the United Kingdom, China and the European Union.
That matters more than it first appears. Kelvin's bakery only sells in Singapore, but his poster is on Instagram, which anyone in the world can see. If he later sells a printed tote bag with the image through an online shop that ships overseas, the question of what rights he holds may be answered under another country's law as well.
For most everyday uses, such as a social post or an internal slide deck, this doesn't cause problems. It starts to matter when you sell something, license it to a client, or want to stop someone else copying it.
One factor comes up again and again in the early decisions and guidance: how much human creative input went into the result.
Compare three versions of Kelvin's poster. In the first, he typed one line, "Mid-Autumn poster for a bakery", and used the first image the tool gave him. In the second, he generated thirty images, picked one, then repainted the mooncake box by hand, added his own photo of the bakery's shopfront and laid out the text himself. In the third, he drew the whole poster himself and used AI only to suggest colour palettes.
Most people would say the third is clearly his work, and the first barely is. The second sits somewhere in between, and that middle ground is exactly where the law is least settled. In places that require human authorship, the parts Kelvin created himself are on firmer ground than the parts the tool produced.
You don't need to know where the legal line falls to act sensibly on this. Knowing that your own input matters is enough to keep track of it, which lesson 4.3, Working rules for AI content at work, comes back to.
When Kelvin asked an assistant who owned his poster, it answered as if the matter were simple. Assistants often do this. They're trained to be helpful and fluent, and a hedge sounds less helpful than a clear answer. AI fundamentals: what it is, how it works, where it fails covers why in lesson 6.1, Why a fluent answer can still be false.
The same goes for a confident blog post, a social media thread, or a well-meaning colleague. If anyone gives you a one-line answer to an AI copyright question, treat it as unreliable unless they're a lawyer who has looked at your specific situation and the country it falls under. This course gives you working knowledge to spot risks and ask good questions. It isn't legal advice, and for anything you plan to sell or license, a lawyer's view is worth paying for.
Kelvin's honest answer to his friend was "I don't know, and I'm not sure anyone does yet." Then he started paying more attention to how much of each piece was his own work. Look back over the AI-assisted things you've made recently with the same eye: the posters, the reports, the captions, the code.
Write down three pieces of AI-assisted work you have produced and note how much of each was your own creative input.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).