Thailand's Electronic Transactions Development Agency (ETDA) released a revised draft of the country's AI Act on July 2, 2026, opening a roughly 30-day public hearing window, according to an analysis published by Bangkok law firm Tilleke & Gibbins. The timing is notable: the release fell squarely inside ETDA's own AI Governance Week 2026, a June 29–July 3 event the agency ran to showcase "AI Governance Practice Center" tools and red-teaming standards. The draft is the latest evolution of a drafting process ETDA and the Ministry of Digital Economy and Society opened in mid-2025, when the agency first put a risk-based framework for high-risk AI — systems capable of affecting the public's rights, freedoms, or safety — out for comment.
What's new in the July draft is the machinery for turning that framework into sector-specific control. A national AI committee would gain the power to designate, by royal decree, entire industries as "high-risk" — with national security, health, environment, energy, telecommunications, transport, and public utilities named as the eligible categories, per Tilleke & Gibbins' review of the text. Once transport is designated, AI systems used in autonomous or driver-assist vehicles — sensor fusion, path planning, object detection — would fall under mandatory risk-management, transparency, and human-oversight duties. The same committee could also designate AI providers serving government or "critical infrastructure" clients as "contract-controlled AI businesses" and require that their data processing occur within Thailand for services deemed of "national importance."
Layered on top is a liability rule stricter than almost anything else in the region: providers and deployers face joint liability for AI-related damage regardless of intent or negligence. The available defenses are narrow — force majeure, the victim's own act or omission, or compliance with an official order. Nothing in the draft, as reported, allows a provider to point to compliance with an international safety standard or a comparative-fault showing as a defense.
Why the Timing Matters for Autonomous Vehicles
This isn't an abstract exercise for the automotive sector. Tesla is among the companies with Full Self-Driving (Supervised) formally listed as pending regulatory approval in Thailand, as shown in a country-by-country rollout map Tesla VP of AI Software Ashok Elluswamy presented at CVPR 2026 in June. FSD depends on continuous camera, radar, and location data, much of which is processed and modeled off-shore for route optimization and neural-network training. If transport is designated high-risk and an FSD-type service is treated as "of national importance," the localization clause would force exactly the kind of in-country processing infrastructure that raises entry costs and slows regulatory approval — at the same moment Thailand is trying to attract EV and AV investment.
The Case for the Rule
The strongest argument for this approach isn't hard to construct. Thailand has no dedicated autonomous-vehicle statute; AV crashes today would be litigated under general tort, product-liability, and traffic law never written with self-driving systems in mind. Proving fault in a software-mediated crash is genuinely hard — the victim usually cannot see the model's decision path, and the information asymmetry between a global AI vendor and an injured motorcyclist in Chiang Mai is real. A no-fault joint-liability rule shifts that evidentiary burden onto the party best positioned to carry it. Data localization, for its part, has a plausible national-security logic: continuous, real-time geospatial and sensor data covering Thai roads, military-adjacent infrastructure, and border areas is a different category of asset than a chatbot log, and Thailand isn't alone in wanting a domestic copy of it.
Why It Still Goes Too Far
But the draft conflates two policies that should be separated. A sector-specific high-risk designation with transparency and human-oversight duties is a defensible, proportionate response to a real accountability gap. Strict, fault-free joint liability with no safe harbor for demonstrated compliance is not — it's a rule that makes the cost of operating in Thailand independent of how safely a company actually operates, which is precisely the wrong incentive for a technology whose entire value proposition is incremental safety improvement over human drivers. Pair that with a data-localization mandate triggered by a discretionary "national importance" finding, and Thailand has built a regime where the compliance bar can rise after a company has already sunk capital into market entry. The EU, by contrast, deliberately withdrew its AI Liability Directive in 2025 rather than layer a bespoke no-fault regime on top of the AI Act, leaving liability to existing product-law and national tort systems. Thailand's ETDA is choosing the opposite path, and doing it through a committee empowered to expand scope by decree rather than through primary legislation debated in parliament.
The 30-day hearing window ETDA opened on July 2 is short for a rule this consequential. If Thai policymakers want AV operators to actually deploy — and the FSD approval queue suggests they do — the liability chapter, not just the high-risk list, deserves the closest scrutiny before this draft hardens into law.