Thailand’s Passport to AI Controversy

In a summary of Prime Minister Anutin Charnvirakul’s speech at China’s 2026 World Artificial Intelligence Conference (WAICO), the Thai government wrote that more than merely adopting AI, Thailand “wants a seat at the table” in determining how the world governs the technology. But far from getting a say in global AI governance, merely adopting AI at home has proven to be difficult.

As the Boston Consulting Group has argued, while full AI sovereignty is an impossible dream for middle powers, they can at least aim to achieve AI resilience: “using, adapting, and governing AI domestically at scale, while minimizing strategic dependencies.” Staying still is not an option, as countries that lag behind in AI adoption may find themselves less competitive and more prone to industrial disruption.

The Thai government has taken a number of steps towards building a degree of this resilience. This includes supporting the building of Thai large-language models such as ThaiLLM and Pathumma and requiring new data centres to provide “tangible benefits” for Thailand such as sharing computing capacity. Internationally, Anutin has refused to take sides in the emerging AI competition between the US and China, saying that Thailand will “remain neutral”.  Although Anutin spoke at World AI Cooperation Organization (WAICO), Thailand did not participate in its establishment.

To increase domestic adoption, the Anutin government launched the “TH-AI Passport” scheme to provide five million Thais with free access to “pro” AI tools. Thais who sign up can access over 30 LLMs (including models that would usually only be accessible with a paid subscription) along with AI skill training courses. Unfortunately, however, the government’s splashiest AI push has been floundering, with several weaknesses in its design and execution, along with controversy over its procurement and assessment.

How the government seeks to assess the success of the scheme is unclear. Opposition politicians have also called for measurable key point indicators.

Firstly, the programme appears to lack a clear objective. Digital Economy and Society (DES) Minister Chaichanok Chidchob has argued that the purpose of the scheme is to increase Thais’ ability to access models that would otherwise be prohibitively expensive, thus reducing the digital divide. At the same time, Chaichanok says that one key goal of the scheme is to increase Thailand’s AI adoption rate, with the government citing a study showing that Thailand’s AI diffusion rate sits at only 10.7 per cent, compared to Singapore at 60.9 per cent and Vietnam at 23.5 per cent (similarly, a Microsoft study found that despite fast growth, Thailand’s AI adoption rate stands at 12.4 per cent, less than half the global average). But these goals conflict with each other: is the AI scheme aimed at new users who have never adopted AI in their workflows, or power users who need access to pro AI tools?

The implementation of the TH-AI Passport raises further questions about the scheme’s target audience. In the author’s testing, just four moderately taxing prompts (such as coding and evaluating a document) with Claude Opus through TH-AI Passport used up a third of the allocated daily credits. The number of credits can be increased through participation in the AI literacy courses, which power users presumably are the least likely to need. There are also monthly quotas and strict length limits on multimedia generation. Users who need state-of-the-art AI tools might not see the TH-AI Passport as a viable choice. The scheme appears to be positioned in a no man’s land: not materially better than existing free options for the regular consumer, while not meeting the bar for power users.

Finally, the procurement process for TH-AI Passport has come under immense scrutiny. Opposition politicians in the People’s Party and Democrat Party have questioned a requirement that the winning contractor advertise the scheme in at least 1,500 convenience stores (a requirement which the DES permanent secretary argues is needed to raise public awareness of the scheme). Additionally, the scheme draws from a digital development fund outside of the usual budgeting process. This allowed the THB1.6 billion (approximately USD48 million) to bypass regular cabinet approval.

How the government seeks to assess the success of the scheme is unclear. Opposition politicians have also called for measurable key point indicators. So far, the government appears to be focused on tracking registration, but the public has been lukewarm towards the scheme, with the latest data at the time of writing showing that only 1.5 million Thais had registered to use TH-AI Passport. Three days after it launched, the DES Ministry stopped reporting the number of daily active users, although Chaichanok later said that there has been 600,000 daily concurrent users. But crucially, even if adoption ticks up, there are no clear indicators being used to show that adoption is translating to productivity.

To properly assess the scheme’s success, the government should commit to measurable indicators of productivity gains. Should it wish to implement similar schemes in the future, it should utilise more transparent procurement and budgetary procedures and ensure that the scheme has a clear objective to begin with. Otherwise, state attempts to build greater AI resilience in Thailand would only rest on shaky foundations.

2026/264