Businesses urged to document human input in AI-generated works to secure copyright
GEORGE TOWN: Malaysia should amend the Copyright Act 1987 to clarify who owns and controls copyright over AI-generated works instead of attempting to address the issue under the proposed Artificial Intelligence (AI) Governance Bill, a lawyer says.
Esther Ong, senior partner at Messrs Esther Ong Tengku Saiful & Sree, said questions of authorship, originality, ownership and copyright exceptions are fundamentally copyright issues that belong under the Copyright Act, which already governs intellectual property (IP) rights and fulfils Malaysia’s obligations under international copyright treaties.
“Embedding copyright rules in a technology-specific statute dates them the moment the technology moves,” she told theSun. Her comments come as the government prepares the AI Governance Bill.
Digital Minister Gobind Singh Deo confirmed on July 13 that the proposed law would regulate AI systems rather than AI-generated content, leaving copyright matters to existing laws.
Ong said the Copyright Act is better suited because it is organised around creative works and legal rights rather than specific technologies, enabling it to remain relevant through successive technological shifts from photocopiers and the internet to smartphones and generative AI.
She added that the Intellectual Property Corporation of Malaysia (MyIPO) should continue administering copyright matters, while the proposed Central AI Authority focuses on AI governance.
She also said the Bill could complement the Copyright Act by requiring greater transparency from AI developers over the datasets used to train their models.
Ong said one of the biggest unresolved questions surrounding AI-generated content is whether the level of human involvement is sufficient to qualify a work for copyright protection.
She added that a prompt itself could qualify as an original literary work if it is sufficiently expressive but that does not automatically extend copyright protection to the AI-generated output.
“The problem isn’t that prompting lacks skill. It plainly involves skill. The problem is control over expression.“
She described prompting alone as “skill without control”, while prompting combined with selection, iteration and modification gives users greater control over the final expression, which is the element copyright law has traditionally protected.
She added that countries are continuing to refine their legal approaches to AI-generated works.
She also said the US Copyright Office has generally adopted a stricter position, while Chinese courts have recognised copyright when there is sufficient human creative control over the final expression.
Ong said Malaysian law similarly focuses on the degree of human contribution in determining originality.
“Even the one country that protected machine-made works on paper has just recommended deleting that protection.“
She added that as AI tools become increasingly sophisticated, AI-generated works also risk reproducing protected creative expression from existing works.
She urged businesses and content creators to reduce those risks by using AI tools trained on licensed datasets when possible, negotiating contractual indemnities, avoiding prompts that reference living artists or protected works, conducting similarity checks before releasing commercially important assets, and retaining records of prompts, iterations and edits.
Ong said primary responsibility for copyright infringement should rest with AI developers because they determine the datasets used to train their models.
“They are better placed to negotiate a single licensing agreement covering all uses, rather than leaving millions of individual users to clear rights they have no way of checking.”
However, she emphasised that users remain responsible for AI-generated content they choose to publish, particularly if they deliberately prompt AI systems to reproduce specific works or imitate protected assets.
She also urged organisations adopting AI to be mindful of broader legal and commercial risks, including uncertainty over copyright protection, infringement claims, inaccurate or fabricated content, authenticity and disclosure issues, and the difficulty of proving human creative contribution.
She highlighted confidentiality breaches as a particular concern for law firms, saying many users overlook restrictions buried in AI platforms’ terms of service before uploading client documents or legally privileged material.
“If the human contribution isn’t documented at creation, it can’t be reconstructed at enforcement and rights that might have existed become unprovable.
“The practical advice is simple: keep the record of the human hand.”
