On 9 February 2026, the Ministry of Science and Technology opened a consultation on guidelines for controlled testing in Decree No. 2026/ND-CP implementing the Law on Artificial Intelligence. The law implemented by this Decree introduces the controlled testing mechanism to allow organisations to trial AI systems that may not yet comply with current laws, offering exemptions and state support to gather evidence for future regulatory adjustments. The Decree specifies that successful controlled testing provides the evidence needed to certify an AI system, tailor its ongoing legal obligations based on its risk level, and guide state agencies in updating the relevant regulations (Article 24). Participants in the trial gain preferential access to State infrastructure and data. To participate in controlled testing, an AI system must be innovative or in development, face a legal barrier that prevents normal deployment, and be accompanied by a comprehensive risk management plan from the applicant (Article 25). Article 26 sets the limits for how long and where an AI can be tested, allowing a maximum of three years per phase with one possible extension, and permitting the geographic scope to range from a single area to several provinces. Participation requires submitting an application via the Ministry's one-stop web portal, after which medium-risk systems benefit from a simplified approval process. The authority to issue testing certificates depends on the geographic scope. Provincial committees handle local tests, while specialised ministries or the Ministry of Science and Technology oversee multi-provincial or complex systems, with defence and security applications entirely exempt from this decree.
Original source