MinLaw and IPOS seek feedback on AI and Singapore’s IP regime

8 September 2026

On 26 August 2026, the Ministry of Law (“MinLaw”) and the Intellectual Property Office of Singapore (“IPOS”) launched a public consultation to seek feedback on the impact of artificial intelligence (“AI”) on Singapore’s intellectual property (“IP”) regime. The consultation closes on 22 October 2026.

AI is increasingly used worldwide to create content, support research and innovation, and assist decision-making. Singapore’s IP regime balances the protection of rights and incentives that support human creativity and investment while enabling innovation for the public good. As AI advances, Singapore’s IP laws and frameworks must remain fit for purpose, trusted, and responsive to emerging developments.

Adopting a co-creation approach to examining AI-related IP issues, MinLaw and IPOS have engaged with rights holders, technology companies, representatives from the creative industries, legal practitioners, academics, and other stakeholders. This public consultation now seeks broader views and perspectives from interested parties on AI-related issues in Singapore’s copyright and patent regimes.

Copyright

Singapore’s Copyright Act 2021 was modernised to address digital-age technological and market developments, including the introduction of a computational data analysis (“CDA”) exception (in sections 243 and 244 of the Copyright Act 2021) which expressly recognises the use of copyright works for machine learning. The consultation seeks to examine how this approach may be refined to further enhance certainty and accountability.

Scope and applications of CDA exception

From MinLaw and IPOS’ engagements and consultations, stakeholders’ views have diverged in two aspects of the scope and applications of the CDA exception. MinLaw and IPOS provide clarification in this regard:

  • Applicability to GenAI and other machine learning applications: The CDA exception is expressly intended to apply to machine learning for all types of technologies, including generative AI (“GenAI”). This is consistent with its technology-neutral design and policy intent to support AI and data-driven innovation in Singapore. Further, the policy intent behind the exception is to cover a broader, rather than narrower, range of technological developments in AI training.
  • Applicability to commercial activities: There is no express limitation on the purpose for which CDA is undertaken (including for commercial purposes).

Legal obligations that safeguard rights owners’ interests

Legal obligations that safeguard rights owners’ interests are key to maintaining the balance struck by the CDA exception. Rights owners may employ legal measures (e.g. contractual obligations) and technical measures (e.g. technological measures) to control access to or use of copyright works. The lawful access safeguard in the CDA exception supports rights owners’ use of such measures: users cannot rely on the exception if they unlawfully access copyright works for AI training purposes.

This consultation concentrates on improving clarity in the key scenarios and perspectives that have surfaced through stakeholder engagements. These key scenarios, outlined in the consultation paper, examine whether the CDA exception applies despite legal or technological measures, for example, whether access in breach of contract is unlawful, whether a term that controls the scope or manner of access is enforceable, and whether a term that seeks to protect the security and stability of online services is enforceable.

MinLaw and IPOS seek feedback on the technical or legal measures that rights owners employ to control access to or use of their copyright works for AI training, whether it is clear from the CDA exception when the exception applies despite rights owners’ use of any such measures, and if it is not clear, whether the issue should be addressed by allowing case law to develop, legislative amendment, or non-legally binding instruments.

Enhanced support for non-legally binding technical measures

At the AI training stage, non-legally binding technical measures include a range of tools, system design choices, and operational practices that enable rights owners to express and/or give effect to preferences regarding access to or use of their works.

MinLaw and IPOS see room to encourage greater respect in the use of copyright works for AI training. This will enable not merely lawful, but responsible, ethical, and trusted AI development. MinLaw and IPOS are of the view that technical safeguards for AI should remain non-legally binding because technology and industry practices evolve faster than legislation can prescribe the safeguards.

MinLaw and IPOS seek views on how non-legally binding technical measures can be developed and adopted in ways that are practical, trusted, and responsive to industry needs.

Strengthening copyright risk management in AI deployment and use

MinLaw and IPOS seek views that may inform the risk assessment and potential liability for copyright infringement in the output phase:

  • Copyright infringement liability when AI generates infringing output: Depending on the facts, an AI developer, AI deployer, and/or end user (i.e. prompter) may be potentially liable for copyright infringement by doing one or more of these infringing acts: (i) unauthorised reproduction; (ii) unauthorised communication to the public; and (iii) authorising infringement. MinLaw and IPOS seek views on how existing legal principles should apply when AI-generated output infringes copyright, including how responsibility should be assessed among AI developers, deployers, and end users. These views may help clarify questions around preventing, detecting, and responding to infringement risks.
  • Technical measures to minimise infringement risks: Singapore does not mandate binding technical measures that target the generation of infringing output, but encourages responsible, ethical, and trusted AI development, deployment and use. MinLaw and IPOS seek views on the technical measures that AI developers, deployers, and users should adopt to minimise the risk of infringement arising from AI output, and the extent to which such measures should be calibrated to the scale and capacity of AI developers, deployers, and users.

Human authorship of AI-assisted works

Under Singapore’s copyright regime, a work that is created with AI’s involvement is only protected by copyright if it is made by a human. The question is how to achieve greater clarity and predictability in differentiating between protectable AI-assisted works and unprotectable AI-generated content. MinLaw and IPOS seek feedback on questions including the following:

  • What existing and emerging use cases of GenAI augment human creativity in the process of creating authorial works?
  • What acts or decisions by creators (i.e. human prompters) demonstrate human creativity when GenAI is involved in creating an authorial work?
  • Should principles or factors be developed to assess the sufficiency of human creativity in such cases? If so, what should these be and should they be developed by the courts, prescribed in legislation, or provided in non-legally binding instruments?
  • How would responses differ in cases where the AI output is executed using an agentic AI system?

Patents

MinLaw and IPOS seek views on how AI is affecting inventorship and patentability with a focus on: (i) assessing human contribution in AI-assisted inventions; and (ii) the effect of AI-generated prior art on patentability.

Assessing human contribution in AI-assisted inventions

Singapore’s position is that only a natural person can be named as an inventor. When AI is involved in generating, evaluating, or refining the technical solution, it may be difficult to distinguish between human contributions that shaped the inventive concept and AI-generated outputs that merely informed, enabled, or accelerated the inventive process. MinLaw and IPOS seek views on:

  • whether Singapore’s current framework, which requires a human to have contributed to the inventive concept in order to qualify as an inventor, is adequate to determine inventorship given the human-AI interactions in innovation and research processes today;
  • the level or type of human contribution that should be required for a person to qualify as an inventor of an AI-assisted invention; and
  • where AI was used in the inventive process, the information that applicants should be required to disclose about the AI model.

Effect of AI-generated prior art on patentability

In Singapore, the requirements of novelty and inventive step are assessed against the “state of the art”, which generally comprises everything made available to the public before the priority date of the invention. In general, anything made available to the public before the priority date may constitute prior art. A claimed invention will not be patentable if it is not new, or if it lacks inventive step if obvious to a person skilled in the art, in light of the prior art.

As GenAI can produce technical descriptions at scale and speed, this may increase the volume of publicly available disclosures and affect later patent searches, applications, and assessments of patentability. MinLaw and IPOS seek views on the challenges such disclosures may present for patent searching, examination, and innovation activities, and views on any practical or policy considerations arising from AI-generated disclosures, as well as whether existing approaches to assessing prior art may require clarification or adaptation to ensure that the patent system continues to support both high-quality patents and innovation.

Reference materials

The following materials are available on the MinLaw website www.mlaw.gov.sg:

More