Thailand’s AI revolution hits turbulence days after its ฿1.6bn platform launches. Paying businesses faced outages, users reported repeated connection failures while Minister Chaichanok pushed free Pro access for free users as the government expanded its dramatic intervention in the commercial AI market.
Thailand’s ฿1.6 billion drive to put professional AI into millions of hands has collided with a chaotic week of malfunctions, inaccurate results and global outages. Digital Economy and Society Minister Chaichanok Chidchob is defending TH-AI Passport while ordering Pro-level access for all who require it and considering free internet as registrations slow. Meanwhile, paying businesses have suffered disruption across AI services critical to daily operations. No link between Thailand’s surge in demand and the international failures has been established, but none has been ruled out. What began as a drive for wider AI access has now seen the government wade directly into an established commercial marketplace, raising questions over capacity, public spending and the real economic costs of its intervention.
Digital Economy and Society Minister Chaichanok Chidchob ended the week defending Thailand’s ฿1.6 billion artificial intelligence experiment. The defence followed an extraordinary week for regular AI users in Thailand. TH-AI Passport opened publicly on August 31. Within days, users were reporting inaccuracies, malfunctions, dropped connections and unexpected problems.
At the same time, major international AI platforms suffered unusual disruption. Several leading services encountered problems within the same narrow period. That timing immediately raised questions about the sudden increase in demand from Thailand. No direct connection has yet been established. However, neither has one been fully ruled out.
The question matters because Thailand deliberately set out to create millions of new AI users. The government is subsidising access to professional-level models developed by major international companies. Many existing commercial customers in Thailand already pay substantial amounts for those same services. For them, the week brought business disruption rather than free access.
Chaichanok says registrations slow as government considers free internet to widen provincial AI access
Chaichanok moved directly into the controversy at Parliament on September 3. He spoke before attending a subcommittee meeting on data centre regulation. TH-AI Passport had then been publicly available for only three days. Yet questions were already mounting over performance, user numbers and the project’s value.
In response, Chaichanok disclosed approximately 500,000 daily active users. Total registrations stood at around 1.3 million to 1.4 million. Earlier active-user numbers covered only the first nine or ten hours after launch. However, the minister said another development concerned him more.
Registration growth was already slowing. Notably, Chaichanok considered that slowdown more serious than possible misuse of the platform model. His ministry had also identified a sharp geographical imbalance. Most users were concentrated in Thailand’s urban areas. Provincial participation remained much weaker.
As a result, the government is already considering a further subsidy. Chaichanok said his ministry would coordinate with the Ministry of Higher Education, Science, Research and Innovation. They would examine affordable or free internet access outside major urban areas. Teachers and students would be among those targeted.
The proposal could also extend to the general public. Its aim is wider access to professional-level artificial intelligence. Accordingly, TH-AI Passport is already expanding beyond subsidised access to private AI models. The government is now considering subsidising the connection needed to reach them.
Chaichanok defends private AI model strategy as users report inaccurate and outdated platform responses
Chaichanok again placed inequality at the centre of the policy. He wants advanced technology available to Thais unable to afford commercial subscriptions. Yet the proposal emerged while the existing programme was already experiencing problems. Reporters had begun testing the system and questioning its results.
Separately, users complained about outdated or inaccurate answers during the programme’s first days. Reporters pressed Chaichanok on those failures. His response revealed important details about the platform’s design. TH-AI Passport does not create its own artificial intelligence models.
Instead, the platform provides access to models developed by private companies. The government acts as an access point rather than a model developer. Consequently, Chaichanok argued that results should match those obtained from the same models elsewhere. The underlying systems remain commercial products.
The government deliberately rejected the more ambitious alternative. Chaichanok said building models capable of surpassing leading international companies would be extremely difficult. He cited OpenAI, Microsoft and Gemini while explaining that decision. Instead, Thailand assembled specialised models through a common platform.
Users can therefore test several leading systems without paying normal commercial charges. That is central to the project’s purpose. However, Chaichanok also identified another reason for disappointing results. Established AI customers often have long working histories inside their existing accounts.
Government orders Pro models as default after blank-slate accounts and free versions weaken user results
Over time, those systems can become personalised around particular users. They retain previous conversations and accumulated information. By contrast, a new TH-AI Passport user begins without that background. Chaichanok described the starting position as a “blank slate”.
That difference can materially change the user experience. An established commercial account may already understand a user’s working preferences. A new account does not possess that knowledge. Even the same underlying model can therefore appear less useful initially.
To address this, Chaichanok said users can provide links to earlier chat histories. The AI can then process existing information. That allows the new system to begin with more context. However, another weakness had already been identified inside TH-AI Passport itself.
The platform was defaulting users to a free model. It was not automatically selecting the highest professional version. Chaichanok acknowledged that problem and ordered a change. The highest-level, or Pro, version should become the default for all users.
Crucially, users should not have to worry about the additional token costs. The government manages those charges through contracts with service providers. Therefore, Thailand is not merely offering basic AI introductions. It intends to provide professional commercial products at public expense.
Chaichanok described the objective directly. “The intention of this project is to let you know which model is right for what you want to use.”
Chaichanok promises Pro ChatGPT access, public usage dashboard and greater state support for AI users
He continued: “You get access to the free, advanced features, and the Pro version of ChatGPT without paying.” Users could then test different products against their own requirements. “By the end of the project, you will know which model you should use.”
For frustrated users, the minister offered practical advice. First, they should confirm that they were using a Pro-level model. They should then try different models according to their needs. Chaichanok expects the experience to improve as the system develops.
On another front, he promised a dashboard showing actual user numbers. The dashboard would allow usage to be checked publicly. It would also answer concerns over inactive registrations or inflated figures. That promise became important after questions emerged about so-called “ghost users”.
Taken together, the minister’s briefing showed how far government involvement has already extended. The state is paying for access and managing token charges. It is selecting default models and combining numerous private services. Now it is considering subsidising internet access as well.
That expansion has another important economic dimension. Thailand already possessed a substantial commercial AI user base before August 31. Businesses, professionals and individual subscribers were paying for premium access. Many had incorporated those systems deeply into daily operations.
For those users, AI is not a government experiment. It is increasingly part of their working infrastructure. Businesses use it for research, coding, analysis, translation and content production. Other users rely upon it for customer services and routine commercial tasks.
Commercial AI users face disruption as businesses rely on paid platforms for daily operations and output
In parallel, professionals have developed complex workflows around particular platforms. Many have accumulated extensive histories inside their accounts. Those histories can become valuable working assets. They also help explain why experienced users can react strongly when services fail.
Commercial customers pay because reliability matters. Speed matters as well. Capacity and continuity also have economic value. When a premium AI platform stops working, the consequences can extend beyond inconvenience.
Employees can lose productive hours. Workflows can stall and deadlines can slip. Scheduled output can be delayed. Staff can also be forced back onto slower manual processes.
This week therefore produced a sharp collision between two different groups. The government was expanding access for people unable to pay. Meanwhile, existing customers were already paying good money for dependable AI services. Many then experienced serious disruption across the same period.
For businesses, the problem was fundamentally different. They were not experimenting with a new free government service. They had already purchased access and organised work around it. Their operations could therefore suffer when the wider AI environment became unstable.
That point places an economic limit on the government’s equality argument. Expanding access benefits the people receiving the subsidy. However, established customers also possess legitimate economic interests. Their position cannot disappear simply because the government enters the market.
Thailand’s AI rollout coincides with global outages as WIRED questions whether failures were linked
At present, those commercial disruptions cannot be attributed to TH-AI Passport as an established fact. No technical evidence supplied has proven that Thailand caused the international failures. Equally, no evidence supplied has ruled out a connection. The timing therefore remains an open question.
Thailand launched the programme publicly on August 31. It immediately began directing a large new audience towards multiple international models. Within the same extraordinary week, major AI platforms experienced serious problems. That sequence does not prove causation.
Still, it attracted attention because the international failures themselves were unusual. US technology magazine WIRED examined simultaneous problems involving OpenAI, Anthropic and xAI. “The issues initially appeared to be linked because they coincided,” WIRED reported.
The publication raised a shared third-party service provider as one possible explanation. Yet the companies later supplied different accounts of their own failures. OpenAI attributed its September 3 disruption to a routing error. xAI pointed elsewhere.
Specifically, xAI blamed an outage at its Memphis compute centre for Grok’s problems. Anthropic said it had identified the cause of its disruption. It also said a fix had been deployed. However, Anthropic declined to provide WIRED with further details.
Global AI outage causes remain divided as Thailand’s new demand leaves technical questions unresolved
Thus, no single public explanation covered the remarkable timing across all three companies. WIRED did not claim Thailand caused those failures. Its reporting is important for another reason. Even specialist international observers initially questioned whether the incidents were linked.
Against that background, Thailand had just created a sharp increase in AI participation. The scale and destination of that demand therefore deserve technical examination. That does not turn a possible relationship into an established one. It simply leaves the question unresolved.
For commercial users, the eventual answer has direct economic significance. If Thailand’s new demand played no role, responsibility lies elsewhere. If some connection emerges, the assessment becomes much more serious. The government programme would then have imposed unintended costs elsewhere.
Again, that outcome has not been established. Yet the possibility cannot be removed simply because it is politically inconvenient. The technical facts must determine the answer. Meanwhile, businesses affected by outages have already borne the disruption.
The episode also exposes competing interests within Thailand’s AI economy. Chaichanok is concentrating on people excluded by cost or inadequate internet access. Established users have already invested money, training and working time. They have built operations around services they purchase.
Government pushes higher AI participation as slowing registrations sharpen concerns over system capacity
Consequently, broader access and commercial reliability are separate economic concerns. The government’s intervention has now brought them into the same policy debate. More importantly, Chaichanok wants participation to increase further. His immediate concern is slowing registrations.
To counter that slowdown, the government is considering cheaper or free internet access. That could bring substantially more users onto TH-AI Passport. Accordingly, capacity questions will become more important as participation expands. Higher demand is not an accidental by-product of this project.
Higher demand is the project’s central purpose. The government wants millions more people using advanced artificial intelligence. Therefore, the supporting systems must cope with the traffic created by success. Registration growth cannot be celebrated while capacity problems are treated separately.
Authentication must remain reliable. Connections must stay stable. Requests must reach the private providers. Those providers must then have sufficient capacity to respond.
The first week showed how many separate components are involved. Some are controlled within Thailand. Others belong to major international technology companies. TH-AI Passport sits between the user and those outside systems.
TH-AI Passport relies on domestic and global systems as outages expose multiple points of possible failure
As part of this structure, failures can occur at several points. A domestic access problem can interrupt service. An international model failure can produce the same result. Routing, authentication and external computing all sit inside the broader chain.
For users, however, those distinctions provide little comfort. The service either works or it does not. Business customers face the same reality. Technical explanations do not recover lost production time.
Likewise, free access for one group does not compensate another group for interrupted commercial operations. That tension gives unusual relevance to a famous warning from former US President Ronald Reagan.
“The nine most terrifying words in the English language are: I’m from the Government, and I’m here to help,” Reagan said.
The remark reflected Reagan’s scepticism towards government intervention in economic activity. Thailand has now entered an established commercial AI market. It has done so explicitly in the name of reducing inequality.
Reagan warning gains relevance as Thailand enters AI market with subsidies and possible free connectivity
The government is spending ฿1.6 billion on that intervention. It is paying for access to products already sold commercially. In addition, it is absorbing token charges for users. It is also deciding which private models appear by default.
Now another ministry could become involved through subsidised internet connections. The stated aim is to reduce the digital divide. Provincial users would gain access closer to that enjoyed in major cities. Nevertheless, each successful expansion would also generate more demand.
That makes capacity an unavoidable part of the equality programme. Reliability is equally important. So is the effect on existing commercial users. None of those questions can be answered simply by counting registrations.
Chaichanok’s 1.3 million to 1.4 million registrations show substantial early interest. His estimate of 500,000 daily active users provides a stronger operating figure. Even so, sustained usage will eventually matter more than either number.
The promised dashboard should provide another measure. It should show whether registrations become genuine activity. It should also show whether users continue returning after the launch period. That will matter as initial curiosity fades.
Government spending, token subsidies and rising demand put reliability and active usage under scrutiny
Notably, experimentation is itself one of Chaichanok’s stated objectives. Users are encouraged to compare different models. They can then identify products suited to particular work or personal needs. The government is effectively financing that discovery process.
Some users may eventually become paying customers of models encountered through the programme. Others may stop using AI when the programme ends. Another group may become accustomed to Pro products they cannot later afford. It is not yet clear which outcome will dominate.
For now, the government remains focused on expansion. Chaichanok regards unequal internet access as a significant obstacle. Hence, the response is to remove another barrier rather than reduce the programme’s reach. That could accelerate demand after registration growth began slowing.
The timing remains difficult. TH-AI Passport opened during an exceptional week for the wider AI industry. At home, users reported inaccurate and outdated answers. The platform also required a change to its default model.
Experienced commercial users faced separate disruption across major international services. Meanwhile, Chaichanok was forced to explain why new users received different results. The “blank slate” problem became part of that explanation. So did the use of free models by default.
Thailand funds AI discovery as users compare models while questions grow over uptake and affordability
In turn, the minister ordered Pro-level access to become standard. He promised transparent usage figures through a dashboard. He also reaffirmed the government’s commitment to paying the associated costs. Those steps considerably deepen the state’s involvement.
The government is no longer merely promoting AI literacy. It is financing usage and directing consumers towards commercial platforms. It is also considering financing connectivity. That creates a much larger operational burden than a conventional education campaign.
Furthermore, the government cannot control every part of the resulting user experience. The underlying models remain private international products. Their computing systems sit outside Thailand’s control. Their outages can therefore affect Thai users regardless of domestic performance.
The reverse question remains unanswered. Thailand has sharply increased access to those external systems. Whether that additional demand affected anything outside Thailand has not been established. Yet the possibility remains, for now, an open question.
The distinction is especially important after WIRED’s reporting. The international failures appeared unusual even before Thailand entered the discussion. OpenAI, Anthropic and xAI encountered problems at almost the same time. Their subsequent explanations were not identical.
Government deepens AI role with Pro access as international outages complicate the national rollout
OpenAI identified a routing error. xAI cited its Memphis compute centre. Anthropic identified its own cause but offered WIRED no further detail. Consequently, the wider sequence remains more complex than one simple outage.
Thailand’s programme therefore landed at a particularly awkward moment. The state had just invited a much larger population into professional AI usage. Existing international services then suffered an exceptional run of disruption. Thai commercial users were caught in the middle.
Those businesses deserve a place in the economic calculation. They already paid for access. They did not require a subsidy. Instead, they required the services purchased for commercial purposes to remain dependable.
Their losses have not been quantified yet. Nevertheless, the mechanisms of disruption are straightforward. Lost hours reduce productivity. Delays can disrupt schedules. Interrupted AI workflows can force additional manual work.
Chaichanok’s equality policy therefore now sits beside a second economic issue. Government-supported newcomers are receiving professional AI without paying directly. Existing subscribers continue paying market rates. Both groups ultimately depend upon much of the same international technology.
OpenAI, xAI and Anthropic give different outage explanations as Thai businesses absorb disruption
That does not make the interests incompatible. However, it makes reliability central to the programme’s success. Expanding access while services fail would undermine the practical value of the subsidy. Commercial users would also continue bearing costs outside the programme.
Another tension lies inside Chaichanok’s own September 3 briefing. He acknowledged system complaints and disappointing results. Yet he said slowing registrations worried him even more. The government’s answer is therefore to increase access further.
That position is clear. Officials want more people using the platform despite a difficult opening week. They also want better models selected automatically. Moreover, they want internet costs reduced for people outside urban centres.
The programme is consequently moving on several fronts at once. Participation must increase. User experience must improve. Pro access must become standard. At the same time, the government must prove that actual usage justifies the expenditure.
The dashboard will therefore become important. Registration totals alone cannot answer questions about cost-effectiveness. Daily activity provides a better measure. Persistent use over several months would provide stronger evidence still.
Chaichanok has put the current active-user figure at roughly 500,000 daily. That is already a substantial audience. If provincial participation rises, the number could increase considerably. The government has explicitly designed the project to achieve that result.
Chaichanok’s equality drive faces reliability test as registrations slow and government widens access
Accordingly, infrastructure planning cannot lag behind recruitment. The platform must cope when people arrive. External services must also cope with the requests reaching them. Otherwise, successful promotion can create its own operational problems.
Thailand’s experience this week has already exposed that risk in practical terms. Users encountered problems almost immediately. Some answers appeared inaccurate while connections were lost and failed. Others differed from established paid accounts. Meanwhile, the wider international AI market was dealing with the most serious outages ever seen, certainly in Thailand.
The problem appeared to be a lack of capacity and the ability to connect worldwide.
Chaichanok has offered explanations for the domestic issues. New accounts lack accumulated personalisation. The default model was not always the strongest available. The underlying AI itself comes from outside providers.
Those explanations answer part of the user experience problem. They do not settle the broader infrastructure question. Nor do they settle the international timing. Both issues remain separate from the government’s stated intention.
The intention is clear enough. Chaichanok wants poorer users to experience advanced AI. He wants provincial communities to gain access comparable to urban areas. He wants users to discover which commercial model best fits their needs.
However, the undertaking now extends well beyond those objectives. The state is committing substantial public money. It is creating demand and paying usage costs. It is also entering an existing commercial market containing established paying customers.
Infrastructure must keep pace as Thailand expands AI access amid domestic problems and global outages
Reagan’s famous warning therefore fits the circumstances without requiring any broader political argument. The government has entered the market promising help. The help comes through subsidies, free access and possibly free connectivity.
Within days, the programme required explanations over malfunctions, inaccurate responses and default model quality. Registration growth also began slowing. At the same time, international platforms suffered a rare cluster of failures.
The commercial side of the market was also disrupted. Existing subscribers had paid for dependable professional services. Their businesses did not stop needing those services because the government wanted to reduce inequality elsewhere.
Thailand’s AI push has therefore moved quickly beyond a simple digital-access programme. It now involves public expenditure, commercial competition, infrastructure capacity and business continuity. The issues emerged within hours of the public launch.
Chaichanok remains committed to pushing ahead. The minister wants registrations growing again. He wants provincial participation increased. He wants Pro-level models available automatically and without concern over token costs.
Thailand’s AI intervention expands beyond access as public spending and commercial pressures intensify
The next test is therefore operational rather than promotional. The government must show that people keep using the system. It must also show that the system remains dependable as participation rises.
At the same time, commercial users will continue judging the same wider AI ecosystem by different standards. They already pay for performance. They expect continuity because their work depends upon it.
The wider international disruption has made that distinction impossible to ignore. Whether Thailand contributed remains unanswered. Whether the week damaged working routines is not. Neither is the fact that the outages caused economic loss.
People’s Party firebrand attacks flagship TH AI project launched this week with corruption agency filing of claims
Thais take to government TH AI project on the first day with ChatGPT reporting outages later on Monday
For Chaichanok, the challenge is now considerably larger than registering millions of users. The government must demonstrate that its ฿1.6 billion intervention works at scale. It must also account for the pressures created by that scale.
Thailand came into the week promising broader AI access in the name of equality. It ended the week explaining failures, user behaviour, model defaults and internet inequality.
Meanwhile, paying businesses were left dealing with interruptions across the same technology market. That is the harder economic story now sitting behind TH-AI Passport.
The government wanted millions more people using advanced artificial intelligence. It has begun delivering them.
Now the systems, contracts and infrastructure must carry the load.
Receive all our stories as they come out on Telegram here
Follow Thai Examiner here
Further reading:
Ministry fails to halt ฿25.4 billion payment order to Hong Kong based firm linked to 1990 contract
Microsoft accepts charges of corrupt payments in Thailand as it makes a deal with US investigators
Provincial agency boss arrested in Mukdahan after trying to extort ฿2 million from a businessman
