Thai AI Passport Glitches Blamed on Wrong Model Selection
Thailand's AI Passport platform has 1.3 million registrations to date, but a Digital Economy minister blamed data glitches and slow adoption on using third-party AI models and high internet costs for disadvantaged users.
Speaking at parliament on September 3, Digital Economy and Society Minister Chaiyachote Chidchob addressed fluctuating user figures for the TH-AI Passport platform. He confirmed roughly 500,000 registered users, explaining that earlier figures cited by opposition lawmakers represented only morning registration numbers from the launch day. The team is updating the total to current figures, with 1.3 million registrations to date.
The minister acknowledged a slowdown in registrations and identified internet costs as a major barrier preventing disadvantaged communities from accessing the free AI service. Most registrations remain concentrated in urban areas. He said the government is working with the Ministry of Higher Education, Science, Research and Innovation to reduce internet costs for students and teachers.
Regarding reported data distortions—such as Thai input producing output in other languages—Chidchob clarified that the TH-AI Passport platform aggregates external AI models rather than developing its own. The platform integrates models from private companies, several of which are free. He acknowledged it is difficult for any government to develop an AI model better than those from OpenAI, Gemini, or other major providers, justifying the use of third-party models.
Chidchob explained that apparent performance differences between free and premium versions stem from machine learning. Each AI model learns from user interactions to better tailor responses over time. When users first switch to the government platform after extended use elsewhere, the system starts fresh and produces different results. He emphasized that the platform should integrate historical data and training methods to improve consistency.
