Thank You

You are now registered for our Rouse Insights Newsletter

What’s happening in AI-related legislation across Asia? Part 1

Published on 25 Jul 2024 | 4 minute read
Singapore, Indonesia and Malaysia – Data centres drive AI forward

Following our webinar on AI legal developments in ASEAN and China, we’ve published a series of three articles looking closely at the different countries. This first piece covers Singapore, Indonesia and Malaysia.

(The second article looks at Vietnam, Thailand and the Philippines, while the third provides an overview of developments in China.)

 

Much of the AI legal development in Singapore and its immediate neighbours is viewed through the lens of data centres.

There are more than 70 such centres in Singapore, which means the country is a prime location for running Large Language Models (LLMs) for machine learning. Copyright owners and AI developers alike have an interest in relevant legal frameworks.

“Singapore is the region’s most forward-thinking country on AI. The government is taking a gentle approach aimed at fostering innovation. It’s the go-to place for the data centres needed to support AI applications,” says Rouse Principal, Kin Wah Chow.

Citing energy and other sustainability concerns, Singapore put a moratorium on new data centres in 2019. While this moratorium has since been lifted, it created a spill-over effect of data centres to the nearby Indonesian island of Batam, and the Malaysian state of Johor. Malaysia has since announced some big data-centre partnerships and plans, including the development of a USD 4.3 billion facility with Nvidia.

When it comes to AI governance, all three countries have drawn up frameworks and ethical guidelines. But compliance is still largely voluntary. At the regional level, the non-binding ASEAN Guide to AI Governance and Ethics was released in March 2024.

“Both Singapore and Malaysia share the same common-law tradition, so there may be some similarities in the application of IP concepts. This is especially relevant for copyright issues when it comes to the question of machine learning,” says Chow.

Defending AI infringement claims

Copyright infringement is one of the main risks for machine learning in the AI sphere. From the AI-developer perspective, local copyright law is a relevant factor in determining where a machine-learning operation is run from.

“Automated data scraping involves sending a high frequency of requests to a website for the content to be scraped. This scraping involves reproductions, and therefore exposure to authorised reproduction under the relevant copyright law. In contrast to regular human browsing, such high frequency scraping is not likely to justify the defence of implied consent,” explains Chow.

For AI developers running machine learning within data centres in Singapore, invoking the Computational Data Analysis exception is a possible defence under Singapore Copyright law. This exception refers to using a computer program to extract and analyse data from a work, or employing the work to enhance the program's functionality.

There are three necessary conditions for invoking the defence:

  1. The use of the third-party work is indeed for computational analysis as its described in the law.
  2. The user had lawful access to the copy of the work that was accessed.
  3. If the copy is an infringing copy, the user was not aware of that.

“The Computational Data Analysis defence is valid only within Singapore's jurisdiction, so AI developers operating data centres in the country may be able to raise it. Terms of use prohibiting data crawling or scraping may not negate the defence, but there are still some uncertainties on the boundary of this exception,” explains Chow.

Content owners should strengthen their terms of use and technological protection measures/digital rights management, as these can possibly counter the computational analysis exception.

“In Indonesia, it’s quite hard to prove infringement from webpage scraping, as the judges are quite legalistic and need to see actual evidence of infringement. They are not likely to arrive at a conclusion of infringement by making inferences, without tangible evidence of infringement. Also, there is no discovery procedure in Indonesia to aid the procurement of evidence.”

“In terms of the legal framework, it may be difficult to sue data centres in Indonesia  for data scraping,” states Chow.

Invoking fair-use defence

Singapore and Malaysia all have articles in their copyright law that can be invoked as part of a fair-use/dealing defence. The factors for fair-use/dealing are similar, but not identical.

The Singapore Court of Appeal (SCA) was asked to consider a case brought about by Global Yellow Pages against an entity called Promedia. The plaintiff claimed that the defendant had infringed its copyright by photocopying and scanning the Yellow Pages to access the data therein. The SCA ruled in favour of the defendant after being asked to consider several fair-use factors.

This case may be a forestate of how the Singapore court could apply the fair-use defence factors in a machine-learning scenario. Factors include the purpose and character of the dealing, the nature of the work or adaptation, the amount and substantiality of the part copied, and the effect of the dealing on the potential market.

“The Singapore Court of Appeal has indicated that if the position of the law is unclear, then the court is likely to lean towards supporting technological advancement,” explains Chow.” Malaysia's position may be influenced by Singapore's permissive approach to fair-dealing defences, but note that the factors are not identical.”

“In Indonesia, there is talk that the government may now issue an AI regulation. We hope that they will consult with industry before doing so, and will not be overly prescriptive in their model. The latest we hear is that that the government there still wants to hold off and get more feedback before they do anything,” concludes Chow.

30% Complete
Principal
+62 21 5080 8157
Principal
+62 21 5080 8157