On 26 August 2026, Singapore's Ministry of Law (MinLaw) and Intellectual Property Office of Singapore (IPOS) opened a public consultation on artificial intelligence and the country's intellectual property regime. Feedback is due by 5pm on 22 October 2026.

This is a consultation, not a change in law. But the questions map closely to the way marketing, communications and creative teams already produce work. That makes the process relevant before the Government reaches any conclusions.

Three copyright questions with operational consequences

The consultation covers copyright and patents. For brand and content teams, three copyright areas deserve attention.

The first is certainty and accountability in AI training. MinLaw and IPOS are asking how Singapore's computational data analysis exception should operate, what counts as lawful access to copyright works, how rights-owner safeguards and contract terms should function, and whether practical non-binding technical measures could help.

The second is risk when AI systems produce infringing output. The paper says that, depending on the facts, an AI developer, deployer and/or end user may potentially be liable under existing copyright principles. It seeks views on how responsibility should be assessed. It also describes minimising output infringement as a shared responsibility for developers, deployers and users within their respective domains.

The third is human contribution in AI-assisted work. Singapore's framework places human authorship at the centre of copyright protection while allowing AI to be used as a tool. The consultation examines how human choices might be understood and evidenced before generation, during selection and arrangement, and through post-generation editing.

The Government has not adopted those factors as a final test. The paper says the significance of prompting is still evolving across jurisdictions and that the Government does not presently express a view on the possible considerations it summarises for consultation.

Style is not the same as protected expression

One distinction in the paper is particularly useful for creative teams. It says that reproducing the general style of a work does not by itself amount to infringement because copyright protects particular expression, not underlying ideas, facts or stylistic elements.

That does not make imitation risk-free. An output may still substantially reproduce protected expression, and an infringing copy shared on a website or social platform may create further issues. The practical point is narrower: a team should assess what an output reproduces, not rely on a vague claim that every stylistic resemblance is either automatically prohibited or automatically safe.

Move provenance out of the footnotes

For marketing teams, the more useful implication is operational. If an organisation cannot reconstruct how an AI-assisted asset was made, it will struggle to assess source risk, explain human contribution or investigate a complaint.

While the consultation is open, teams can map where AI enters their workflow and decide which records are proportionate to keep. Depending on the work, those records might include source materials and licences; the model, tool and vendor used; the human brief and relevant prompts; generated alternatives; material edits; the approver; and the final publication decision.

Keeping those records does not automatically create copyright protection or eliminate liability. It does something more basic: it makes the process reviewable. That is useful whether the final policy response involves legislation, guidance, industry measures or no immediate change.

Contracts and approval need separate questions

AI-enabled work often crosses several organisations: a client, agency, production partner, software vendor and model provider. Brand teams should therefore ask where contracts address source rights, warranties, disclosure of AI use, record retention, incident response, indemnities and final approval responsibility. These are review questions, not a statement of what Singapore law currently requires.

Approval processes also need more than a single aesthetic check. A provenance and source-risk review asks where inputs came from, what the tool produced and what people changed. A brand and quality review asks whether the work is accurate, effective and fit to publish. Combining both into one hurried sign-off makes it easier for each to be assumed rather than completed.

Organisations with substantial experience of AI-assisted creative or content production may also have useful operational evidence to submit before the consultation closes. Participation is optional, but practical evidence about workable records and real vendor chains can help turn an abstract policy question into an informed one.

What Synthminds is watching

The unresolved issues are material: the consultation outcome, any later legislative or non-binding measures, and how Singapore courts apply existing principles to particular workflows. Until then, the disciplined response is not to predict the law. It is to make AI-assisted production traceable enough that the organisation can understand and defend its own decisions.

This Briefing is general information and does not constitute legal advice.

Sources