Can AI-generated content be copyrighted, and is AI training legal?

This is one of the most debated legal questions about AI, and the law is still evolving.

Two separate questions:

1. Is it legal to train AI on copyrighted material?

• AI models are trained on huge amounts of text, images and music, much of it copyrighted.
• Companies often argue this is allowed as "fair use" or a similar exception, because the model learns patterns rather than copying works.
• Authors, artists, musicians and news organisations argue it uses their work without permission or payment.
• Many lawsuits are under way in the US, Europe and India. In India, the news agency ANI sued OpenAI in the Delhi High Court in late 2024 over use of its content, and other publishers joined. Some AI firms have also signed licensing deals with publishers.

2. Who owns content that AI creates?

• Many countries, including the US, say copyright needs human authorship. Content made purely by AI, with no meaningful human creativity, may not be protected.
• Where a human makes substantial creative choices, such as selecting, arranging and editing, those parts may be protected.
• India's Copyright Act, 1957 was not written with AI in mind, and the government has been consulting on whether changes are needed.

Practical advice: if you use generative AI for business, check the tool's terms of use, avoid copying recognisable works or styles, and add genuine human work.