Flint advises international businesses and investors on policy, politics, regulation and competition.
Flint advises international businesses and investors on policy, politics, regulation and competition.
Quick links
Sectors
Follow Us
AI-Pac is Flint’s digest providing our take on major AI policy and regulatory developments in Asia-Pacific. In a fast-moving environment, this newsletter is designed to help you stay on top of the regulatory agenda and consider what it means for your business. Please see here for our first ten blogs, spanning 2024 and 2025. In each blog, we set out:
Between August and October 2025, there was a marked shift in various countries in the Asia-Pacific region towards horizontal or sector-specific regulation of AI, with continued emphasis on domestic AI infrastructure development. Countries also progressed pending horizontal AI legislative proposals. In this digest:
If you would like to discuss how AI policy and regulation in Asia-Pacific will impact your business, please get in touch with Flint’s team here.
On 22 October 2025, India’s Ministry of Electronics and Information Technology released proposed amendments for stakeholder comments, introducing labelling requirements for AI-generated content. Platforms allowing users to create or modify such content have to ensure that such information is labelled or embedded with permanent unique metadata. There are also additional obligations imposed on “significant social media intermediaries” to obtain and verify a user declaration on whether the information uploaded by them is AI-generated.
The virality of deepfake content online has been a growing concern for the Indian government, with particular concerns in relation to upcoming state elections. India’s election watchdog issued an advisory on 23 October 2025, directing all political parties and candidates to label AI-generated content, and remove such content on official political party social media handles within 3 hours of being notified. There have also been a growing number of court cases initiated by Indian celebrities seeking takedown of misleading AI-generated content using their images.
The amendments mark a significant shift towards hard-law AI regulation in India, even as the country eschews broad-based EU-style regulation for now. There are no exemptions for harmless AI-generated content or protected speech such as parody or satire, which can result in the labelling requirement applying to a vast range of content on social media platforms or content generation platforms like Sora. Failure to comply with these requirements can result in platforms losing liability protection in relation to third-party content. Affected businesses should consider highlighting practical challenges around the design of the rules and proposing targeted amendments during the ongoing consultation process.
Labelling requirements for AI-generated content have become an increasingly common feature of AI regulation in the Asia-Pacific region. China has already enacted such requirements, while both Vietnam and Korea’s proposed AI laws (discussed in more detail below) include labelling requirements. Other governments in the region will monitor the implementation of India’s new requirements closely.
Since its re-election in May 2025, Australia’s government has been reassessing its approach to new, economy-wide AI regulation. A consultation in late 2024 had proposed to introduce mandatory guardrails in high-risk settings. However, in recent months, Australia’s messaging has evolved: in May, Treasurer Jim Chalmers described AI as “a key part of our productivity agenda”, while in September, Digital Economy Minister, Andrew Charlton, indicated that Australia intends to regulate AI “as much as necessary and as little as possible”. A Treasury report released in October noted that Australia’s existing consumer and competition laws can sufficiently address concerns arising from AI products.
Having previously been a key proponent of AI regulation in Asia-Pacific, Australia’s shift in stance on AI regulation has implications for the wider region. Ministers are acutely aware of slow productivity growth in Australia and see promoting AI adoption as a key tool to addressing this. They are also conscious of a broader shift in global attitudes to AI regulation: the US has warned against new regulation around the world that targets US companies, while the UK - which Australia monitors closely - has delayed its own plans for horizontal AI regulation.
At the same time, Australia’s Labor government still seeks to strike a careful balance between promotion of AI and domestic pressures to regulate to protect Australian workers and creative industries. The government is managing demands from its traditional trade union allies to go further on regulation, with the Australian Council of Trade Unions (ACTU) advocating for a new national AI authority to oversee "mandatory enforceable agreements" in every workplace, to require consulting of staff before technology is rolled out. State governments are also influencing the debate: New South Wales’s Labor government introduced a bill seeking to limit the use of AI in the workplace in August.
Meanwhile, responding to pressures from Australia’s creative sector, Australia’s Attorney General said in an official statement that a text and data mining (TDM) exception will not be introduced under Australian copyright law, despite a recommendation from the Productivity Commission in August. A broad coalition of trade unions and creative sector stakeholders had vocally opposed the Productivity Commission’s proposal, reflecting similar mobilisations of the sector in the UK and elsewhere.
There have also been specific instances of regulatory interventions on AI-related issues. On 9 September, Australia’s eSafety Commissioner registered six industry codes to implement online safety measures for children, including for AI companion chatbots. The Commissioner also directed four popular AI companion providers, including Character.AI, to provide information about their compliance with the Online Safety Act.
Altogether, these developments show that, even as Australian ministers seek to deprioritise broad-based new AI regulation, the country will continue to be a hotspot for debate on the societal impact of the technology.
On 8 October, Digital Minister Gobind Singh Deo told parliament that Malaysia is moving beyond voluntary AI guidelines towards a risk-based AI law. The proposed AI law will impose stricter obligations for high-risk technologies i.e. those “with the potential to compromise human rights, security or information integrity”. It would define and categorise AI-related harms, introduce incident reporting requirements, and guide remediation efforts.
For now, the thinking on new, horizontal regulation in the Malaysian government is at a relatively early, conceptual stage. A National AI Action Plan due in December 2025 may include more details on regulation. In the meantime, businesses should consider engaging with the National AI Office - which is leading much of the work - to highlight the importance of targeted and proportionate approaches to new regulation.
Through its ASEAN chairmanship, Malaysia has also progressed work aimed at encouraging greater coordination of approaches to AI governance between ASEAN nations. On the sidelines of the ASEAN Summit, Digital Minister, Gobind Singh Deo, said that the ASEAN AI Safety Network (ASEAN AI Safe) will launch with a secretariat in Kuala Lumpur in early 2026.
In parallel, the debate on sovereign AI has also gathered pace in Malaysia. Malaysian conglomerate, YTL, launched Malaysia’s first homegrown AI model, ILMU. This ignited a conversation about “AI sovereignty” and what that entails. Prominent voices within government pushed to define sovereignty as not requiring the data be hosted locally, but that the model that runs AI solutions for government should be owned locally. Malaysia also announced an investment in its 2026 Budget of RM 2 billion in a Sovereign AI Cloud. In the context of trade tensions and heightened concerns about strategic dependence on foreign companies, Malaysian ministers have linked the concept of sovereign cloud to national sovereignty, arguing that this is necessary to protect sensitive and classified information.
The debate around what AI sovereignty should mean in practice for Malaysia will continue, as the government explores trade-offs around how the pursuit of “full-stack” sovereignty could impact on the pace of AI adoption in the country.
In September 2025, Vietnam’s Ministry of Science and Technology (MoST) released a draft AI law for public consultation until 20 October, with plans for it to take effect in January 2026. The law categorises risks into four levels—unacceptable, high, medium, and low—similar to the EU framework; however, the definitions of these tiers are ambiguous and afford the government significant interpretive latitude, particularly in situations where AI systems may threaten or "sabotage" political interests. Violations may result in a range of penalties, including administrative sanctions, criminal prosecution, monetary fines, temporary suspension, or permanent ban of the AI deployment.
Vietnamese ministers stress three key aspects driving Vietnam’s approach to AI policy: firstly, it should be open enough to attract private investment from companies, including foreign companies (recent Qualcomm investment in the country is viewed as a positive example); secondly, it should be sovereign to ensure continued control over Vietnamese citizens’ data; thirdly, it should be responsible in seeking to avoid technological dependence on foreign powers, particularly given concerns about the impact of tariffs on Vietnam’s export-oriented economy.
Businesses should keep these broader considerations in mind as they determine their approach to engaging on AI regulation in Vietnam. The breadth of the AI law’s provisions means there will be significant discretion around how it is applied in practice, creating opportunities for businesses to highlight the need for a proportionate approach to implementation to ensure Vietnam can benefit from AI. Ministers recognise the potential for AI to drive economic growth, with a previous report from the National Innovation Centre (NIC) (operating under the Ministry of Finance (MoF)) estimating that AI could add up to $130 billion to Vietnam’s economy by 2040.
In December 2024, Korea was the first Asia-Pacific country to pass horizontal AI regulation with its AI Basic Act. Ahead of the Basic Act coming into force in January 2026, on 8 September 2025, the Ministry of Science and IT (MSIT) released a draft Enforcement Decree. The draft Decree outlines obligations, including - echoing developments in India - mandatory watermarking of AI-generated content. Significantly, MSIT did propose a one-year grace period for companies subject to a fine.
Ahead of this Decree, there had been calls in Korea to impose a moratorium on the regulatory provisions of the AI Basic Act, echoing similar demands in the EU. Ultimately, the government decided against a moratorium, though stressed that it is pursuing a “loose regulatory framework” in order to allay domestic concerns about the impact of regulation on Korea’s AI competitiveness and to pre-empt any potential US opposition. Nevertheless, industry has still expressed concern about the extensive self-assessments, documentation, and risk reporting required for high-impact AI.
Despite softening its approach to implementation, Korea’s horizontal AI regulation will set an important precedent. Other governments around the region will be watching closely to see whether the regulation can be effective in mitigating against AI safety risks without impacting on adoption of the technology.
This blog was written by Arpit Gupta and Ewan Lusty, with the input of Flint’s Senior Adviser network based across Asia-Pacific. Ewan and Arpit are both based in Singapore. Ewan has been supporting Flint clients on a range of digital policy issues for over four years, prior to which he advised UK government ministers on online safety regulation. Arpit is a technology law and policy professional with prior experience working with California’s privacy regulator and India’s Ministry of Electronics and IT.