Vietnam's Draft AI Safety Rule for CNC Machines Requires Dual Certification

Ho Chi Minh City, May 22, 2026 — Vietnam’s National Institute of Standards and Industrial Research (VSTL) published the draft Technical Specification for Safety of Intelligent Machine Tools on May 22, 2026. The regulation introduces a mandatory dual-certification requirement for AI-integrated CNC equipment entering or operating in Vietnam — marking a significant tightening of technical market access rules for industrial automation systems.

Event Overview

On May 22, 2026, VSTL released the draft Technical Specification for Safety of Intelligent Machine Tools, stipulating that all CNC equipment incorporating AI functions—including adaptive control, vision-based positioning, and predictive maintenance—must obtain concurrent certification from both VSTL (through its domestic AI safety testing framework) and an internationally recognized body accredited to UL/IEC 62443-4-2 standards. Neither certificate may substitute for the other. Public consultation on the draft runs until June 20, 2026.

Industries Affected

Direct Exporters & Importers
Companies exporting AI-enabled CNC machines to Vietnam will face extended lead times and higher compliance costs due to the need for parallel certification pathways. Unlike previous single-accreditation regimes, this rule eliminates mutual recognition options — meaning exporters must now allocate separate budgets and timelines for VSTL’s local test protocol and UL/IEC 62443-4-2 validation. Customs clearance delays are likely during the transition period, especially for models previously certified only under international schemes.

Raw Material & Component Suppliers
Suppliers providing AI modules (e.g., edge inference chips, real-time vision processors, or embedded ML firmware) to CNC OEMs must now ensure their subcomponents support traceable, auditable safety lifecycles aligned with both VSTL’s AI-specific test cases and IEC 62443-4-2’s secure development lifecycle (SDLC) requirements. This increases documentation burdens and may necessitate joint testing agreements with integrators — particularly where firmware-level security controls (e.g., model integrity verification, runtime anomaly detection) fall outside standard component certifications.

Machine Tool Manufacturers & Integrators
Domestic and foreign manufacturers assembling AI-CNC systems in Vietnam — or supplying turnkey production lines — must revalidate entire system architectures, not just individual subsystems. VSTL’s draft emphasizes end-to-end AI safety assurance, including data provenance, decision transparency, and fail-safe behavior under adversarial input conditions. This implies redesign efforts for legacy AI modules lacking explainability interfaces or secure update mechanisms.

Supply Chain Service Providers
Testing laboratories, certification consultants, and conformity assessment bodies active in Vietnam’s industrial sector will need to expand capacity for AI-specific evaluation — particularly in areas such as adversarial robustness testing, training data bias auditing, and real-time inference monitoring. Those without dual accreditation (VSTL-authorized + IEC 62443-4-2 competent) risk losing engagement opportunities, especially from multinational OEMs seeking one-stop compliance support.

Key Focus Areas & Recommended Actions

Review Product Architecture Against Both Certification Criteria

Manufacturers should conduct a gap analysis mapping current AI functionalities (e.g., closed-loop adaptive feed control, camera-guided tool calibration) against VSTL’s emerging test scenarios — which reportedly include scenario-based stress tests for misclassification under low-light or occluded conditions — and UL/IEC 62443-4-2’s requirements for secure software updates and privilege separation.

Engage Early with VSTL-Authorized Labs

Given limited local capacity for AI safety testing, companies are advised to initiate pre-submission consultations with VSTL-designated laboratories before June 20, 2026. Early engagement may inform interpretation of ambiguous clauses — such as whether cloud-connected AI analytics (hosted offshore) fall within scope — and help shape final regulatory language.

Update Technical Documentation for Traceability

The draft requires full traceability from AI model training datasets to deployed inference logic. Firms should revise design history files, risk assessments, and safety manuals to explicitly document data sourcing, preprocessing steps, validation metrics, and human-in-the-loop fallback protocols — all of which are subject to audit under both certification schemes.

Editorial Perspective / Industry Observation

Analysis shows this is not merely a procedural upgrade but signals Vietnam’s strategic pivot toward sovereign AI governance in critical infrastructure domains. Observably, VSTL’s insistence on local AI testing — rather than accepting equivalence with ISO/IEC 42001 or NIST AI RMF — reflects growing emphasis on contextual risk awareness: e.g., how AI behaviors interact with local power grid instability, workshop environmental variability, or operator skill profiles. From an industry perspective, the dual-certification mandate is better understood as a de facto localization lever — incentivizing foreign vendors to establish in-country AI validation partnerships or even regional R&D hubs. Current more relevant concern is whether VSTL will publish detailed test methodologies prior to finalization — as opacity here could delay implementation beyond the projected 2027 enforcement timeline.

Conclusion

This draft represents a calibrated yet consequential step in Vietnam’s industrial digitalization policy — balancing openness to advanced automation with proactive risk containment. Rather than signaling market closure, it underscores the increasing centrality of verifiable AI trustworthiness in global manufacturing supply chains. For stakeholders, the immediate imperative is not compliance avoidance, but structured readiness: aligning engineering practices, documentation systems, and partner networks with converging safety expectations across jurisdictions.

Source Attribution & Ongoing Monitoring

Official source: Vietnam Standard and Industrial Research Institute (VSTL), Draft Technical Specification for Safety of Intelligent Machine Tools, published May 22, 2026; public consultation open until June 20, 2026. Available at: https://www.vstl.gov.vn/en/regulatory-drafts.
Additional reference: UL Solutions’ guidance on IEC 62443-4-2 implementation (2025 edition).
Areas requiring continued observation: Final scope definition for ‘AI functionality’ (e.g., whether basic PID tuning with online parameter adjustment qualifies), official release date of VSTL’s AI test protocol, and potential alignment pathways with ASEAN harmonization initiatives currently under discussion at the ASEAN Centre for Energy.