AI and copyright: who owns an image made by an AI?

An AI made this image in seconds. So who owns it? Whoever typed the prompt, the software company… or nobody? In 1710, a British law, the Statute of Anne, gave authors themselves an exclusive right to print and sell their books. In 1886, the Berne Convention set up protection for literary and artistic works across several countries; it is still in force. In 2019, a European directive allowed text and data mining, unless rightholders opt out. Three questions: the work, the invention, the data. One: the work. Copyright protects human creation: in EU terms, “the author's own intellectual creation”. In January 2025, the US Copyright Office said a prompt alone isn't enough: only human input, like selecting, arranging, editing, is protected. In 2026, the Supreme Court declined to reopen the issue. Two: inventions. A researcher listed his AI, DABUS, as inventor on a patent. The European Patent Office said no in 2021, the UK Supreme Court in 2023: an inventor must be a person. Three: the data. Models learn by reading millions of works. In Europe, that's allowed, unless the author opts out. Since August 2025, the AI Act requires general-purpose AI providers to respect those opt-outs, and to publish a summary of their training data. Reminder: this is the law as of October 2026, not legal advice; every case depends on the country and the facts. At KOAEE, the assistant uses the business's own content: its texts, its photos. At a restaurant, the photos credit their photographer. At an insurer, on a contract question, it reminds you to check with staff. To understand AI in 150 seconds, subscribe to the channel. The AI that stays between you. Try it at koaee.ai.