Controversy surrounds the TH-AI Passport distributing AI Pro to 5 million Thais. Dr. Rueabin points out value concerns after trials suggest it may not be truly Pro or premium level, with worries that foreigners may be collecting big data instead.
Today (11 Apr 2026), Mr. Chaiyachon Chidchob, Minister of Digital Economy and Society (DES), together with Ms. Nan Boonthida Somchai, Deputy Minister of DES, Mr. Jeset Thaiset, Deputy Minister of the Interior, Mr. Patchara Anantasillap, Permanent Secretary of DES, DES executives, and related parties gathered to listen and exchange views to guide the development of the TH-AI Passport project. This collaboration includes government, private sector, educational institutions, experts, and interested citizens, held at the main conference room, 2nd floor, DES building (Building C), Chaeng Watthana Government Complex, Bangkok.
Following questions about its value, Dr. Thamthee Sukchotirat, known as Dr. Rueabin, a Big Data expert who participated in the discussion, analyzed the TH-AI Passport project's specs and performance based on current project information. The IT side raised key concerns about the technology's capability level.
They worry that the specified specs may not be "Pro" compared to AI commonly used today, whether free or paid versions. This has led to criticism that the project's specs may not match Pro versions of AI technology, or might even perform worse than freely available versions.
This issue will be proven through actual use. It is expected that once launched, the media and users will test features like photo uploading, editing, speech drafting, or document checking to see if the project's performance truly reaches a "Pro" level.
From trials with the project's Demo version at previous events, it was found that although it claims to use the same "brain," its results differ from the AI Pro version I use. The Demo answered only three questions, while my AI Pro could answer ten in detail. Technically, this performance gap might be due to plugins, code differences, and local data management. Handling data locally before sending it may reduce quality compared to direct processing at the source.
The issue of where data is stored and processed has become a major debate. Microsoft's stance at the meeting was that data would not be used to train AI further but they refused to answer whether data is stored or processed within Thailand, leaving the meeting immediately after their statement.
Meanwhile, another agency admitted that data is indeed sent abroad for processing because the main processing unit or "brain" of the AI is located there. For clarity, the expert explained the data process in three parts:
1. Logging in and typing the questions.
2. Sending the questions to be processed by the "brain."
3. Using the data to further develop or train the AI (Training).
Public trust in government platforms: Initial surveys indicate most Thais do not trust government platforms to store personal data. Considering past incidents of Thai data leaks, the government's response should be to apologize and improve. However, the current climate involves suing whistleblowers who expose data leaks while data continues to leak, causing people to fear using Thai platforms.
Recommendations for value and fraud reduction: To address these issues, I used my own AI Pro to draft ten guidelines aimed at ensuring the project is worth taxpayers' money, effective, fair, and reduces corruption risks. Two of these ten guidelines have been approved by the Deputy Permanent Secretary for practical implementation. Additionally, I called for public disclosure of contracts and platform details so citizens can verify the platform's appearance and quality.
I affirm that I do not want to cancel the project and have proposed contract amendments using "contract appendices" to set additional standards and conditions aligned with relevant regulations.
Ultimately, this project should serve as a case study for future AI or digital government projects to reduce vulnerabilities in performance, taxation, fairness, and corruption. I also suggest officials use AI Pro for auditing and writing tasks to increase efficiency and reduce long-term corruption suspicions.
They are as follows:
1. Open real cost accounting; unused funds must be refunded (Open-book & true-up).
Proposal: Contractors must disclose actual per-user costs (rights/Token/cloud). If actual use is less than estimated, the difference must be returned to the state.
Benefit: The state and citizens retain tax money without leakage as excess profit, closing the "profit margin" loophole questioned due to the lack of Token caps in the TOR.
This is possible under Section 97 (state benefits protected) and the value-for-money principle in Section 8.
2. Payment based on actual performance, not just report submission, plus withholding final payment.
Proposal: Instead of paying 20-20-20-20-20 (40% paid within 90 days for just documentation), reduce initial payments, withhold 10–15% as a performance guarantee paid after one year upon meeting criteria, and tie payments to active users, not just registrations.
Benefit: The state and citizens avoid full payment if real users fall below the 5 million target; the state retains negotiation leverage until the end.
This is possible under Section 97 and the value-for-money principle in Section 8.
3. Remove conditions requiring the private sector to cover travel and overseas study costs for government officials.
Proposal: Flight tickets, accommodation, per diem, and overseas training expenses for officials (including inspection committees) should be paid from government budgets, not contractors.
Benefit: Inspectors remain impartial and independent, closing conflicts of interest.
This avoids violations of Sections 127/128 of the 2018 Anti-Corruption Act prohibiting state officials from benefiting from contractors.
4. Remove or revise advertising media items that may favor a single party.
Proposal: Cancel exclusive digital screen contracts in convenience stores covering 6,000 screens/1,500 branches (held by a single rights holder). Allow alternative media use and pay based on actual receipts without added profits.
Benefit: Funds focus more on AI and curriculum development, addressing concerns of media group favoritism.
This aligns with Section 97 and fair competition principles in Section 8.
5. Curriculum, platform, and code must become state property.
Proposal: Expand ownership clauses (originally only covering documents in TOR clause 15.1) to include 96 course modules, media, AI agent systems, and source code as permanent state assets without extra licensing fees.
Benefit: After investing 1.6 billion baht, the state retains assets for ongoing use rather than losing them after one year.
Supported by Section 97 (state benefit) and value-for-money principle in Section 8.
6. Disclose subcontractors and actual shareholders; allow the Office of the Auditor General (OAG) to inspect.
Proposal: Publish names of subcontractors, joint venture members, work shares, and beneficial owners. Allow OAG to audit accounts and receipts throughout the contract term plus five years.
Benefit: Enables real oversight and prevents work being reassigned to cronies.
Enforced by the Integrity Pact (Sections 17–19) and OAG’s audit authority.
7. Review pricing with independent appraisers.
Proposal: Use independent mediators to compare AI rights/cloud market prices and reduce overpriced items.
Benefit: Addresses concerns that "the company set the median price and won by bidding only 1.5% below it."
Possible under Section 97 (state’s interest protection).
8. Increase penalties to be truly impactful and fix conflicting Service Level Agreement (SLA) criteria.
Proposal: Raise tiered fines (originally 0.001% per hour ≈ 16,500 baht, considered too low) and reconcile conflicting standards (216 minutes downtime vs. SLA 99.95%) by adopting the stricter standard.
Benefit: Reduces system downtime, ensuring users receive real quality service.
Supported by Section 97 (state/public interest).
9. Strictly control Thai data from leaving the country.
Proposal: Require contracts with all model owners forbidding using prompts/data to train AI (no-training), specify processing location, and submit Data Protection Impact Assessment (DPIA) reports before service launch.
Benefit: The promise "data stays in Thailand" becomes a binding obligation, not just marketing talk.
Enforced under the Personal Data Protection Act (PDPA) 2019, Sections 28/29.
10. Ensure continuity after one year and return deliverables to the public.
Proposal: Before deleting data within 30 days, deliver data, curricula, and certificates in reusable formats; allow users to download their work and learning history; provide renewal options at pre-agreed prices.
Benefit: Citizens do not lose their work after rights expire; the state avoids vendor lock-in.
Supported by Section 97 and the value-for-money principle in Section 8.