In 1981, photographer Lynn Goldsmith took a black-and-white portrait of the music star, Prince. She later licensed it to the magazine Vanity Fair for US$400, allowing the magazine to use it as a reference for an illustration. The magazine then commissioned iconic artist Andy Warhol to create a stylised version for a magazine cover.
What Goldsmith did not know was that Warhol went beyond the commission – he created 16 additional silkscreen prints based on her photo, without her permission.
Years later, following Prince’s death in 2016, another magazine Condé Nast published one of Warhol’s images, paying the Warhol Foundation but not compensating Goldsmith. She sued, arguing that Warhol’s use of her work without permission infringed on her copyright.
In 2023, the United States Supreme Court ruled in favour of Goldsmith. The key reason? Warhol’s version, while stylistically different, still served the same fundamental purpose – a portrait of Prince. The court found that adding an artistic touch wasn’t enough to qualify as “transformative use” under copyright law.
This case has sparked a broader conversation in the creative industry, especially in the age of generative AI (GenAI), said Karen Robinson, Adobe’s senior vice-president and deputy general counsel.
As AI tools generate new images from existing visuals, similar copyright debates will emerge, she noted, in an interview during the Adobe Summit held in Las Vegas in March.
There is now a growing realisation among governments and artists that a collective approach is needed to address the legal and ethical implications of AI-generated content.
Laws must catch up, Robinson said, pointing to a crucial legal question: When is an AI-generated image different enough to be considered original and copyrightable?
She gave an example. A simple one-line prompt like “create a unicorn sitting under a rainbow” may not be complex enough to claim copyright. But if an artist crafts a highly detailed prompt, shaping the AI’s output into a unique artistic vision, then the final image could be considered copyrightable.
“The emphasis should be on how transformative the AI-generated artwork is compared to the original,” she said.
She added: “The digital marketing industry and AI technology are evolving rapidly, and right now, everyone is just guessing what the rules will be. We need clear guidelines that balance innovation with respect for creative rights.”
When Is AI-generated art copyrightable?
In the US, lawmakers are beginning to act. A proposed bill, the Preventing Abuse of Digital Replicas Act (PADRA), introduced in December 2024, aims to establish legal protections against the misuse of AI-generated content. Adobe has voiced its support, advocating for responsible AI use and safeguards against digital replicas being exploited.
For businesses, the copyright situation is more straightforward, said Robinson. Those that generate AI visuals using their own proprietary data have clear ownership.
However, when creative agencies produce AI-driven marketing campaigns, they must secure proper consent from clients – not just for copyright compliance, but also for privacy regulations.
As AI reshapes creative industries, Robinson emphasised the need for education. “Artists and businesses need to understand how AI works and how it impacts them legally,” she stressed. “Getting this right will allow us to embrace AI while protecting original creativity.”
As of April 2025, Singapore has not announced any new amendments to its copyright laws specifically addressing generative AI (GenAI) visuals. However, the existing legal framework includes provisions relevant to AI-generated content.
In 2021, Singapore introduced the Computational Data Analysis (CDA) exception under Section 244 of the Copyright Act. This provision allows copyrighted works to be used for purposes such as data analysis and machine learning, provided certain conditions are met, including the requirement for lawful access to the original material.
Regarding the copyrightability of AI-generated works, Singaporean law currently stipulates that only works with identifiable human authors can receive copyright protection. This means that content produced solely by AI, without human intervention, is not eligible for copyright.
While there are ongoing discussions about the implications of GenAI on copyright law, no official amendments have been made to specifically address AI-generated visuals in Singapore.
In the past year, Singaporean writers have pushed back against government efforts to use their works to train a multilingual large language model (LLM), citing a lack of clarity on compensation and the intended use of their material.
Beyond monetisation concerns, observers say that excluding Singapore-based writing – rich in the country’s values and culture – could lead to LLMs generating responses that lack local context and nuance.
