Compensation for AI related damages

From 2026, enterprises may be liable for damages caused by artificial intelligence (AI) even without violating the Law. Is this a real risk?

The Law on Artificial Intelligence 2025 (“AI Law 2025”) establishes, for the first time, a dedicated legal framework governing regulatory inspection, administrative enforcement, and civil liability for damages caused by AI systems. The AI Law 2025 does not recognize AI as a legal person capable of bearing liability; instead, it allocates liability based on the degree of human and organizational control over risks arising within the AI value chain.

This approach has significant implications for enterprises deploying and operating AI systems. From 2026, even where an AI system is fully managed in compliance with applicable laws, enterprises may still incur liability to compensate for damages if such damages occur.

In this article, BLawyers Vietnam analyzes 04 key issues relating to compensation mechanisms and legal liability arising from AI systems that organizations and individuals operating in the AI sector should carefully consider when investing in and operating AI in Vietnam from early 2026.

1. How does AI liability shift from a “fault based” regime to a “risk allocation” regime?

Article 29 of the AI Law 2025 introduces a groundbreaking principle under which liability for damages may arise even in the absence of any compliance violation by the enterprise.

Accordingly, where a high-risk AI system (defined as a system capable of causing significant harm to life, health, lawful rights and interests of organizations and individuals, national interests, public interests, or national security) is managed, operated, and used in full compliance with applicable legal requirements but nevertheless causes damage, the deployer shall still be liable to compensate the injured party.

This approach demonstrates that lawmakers have shifted their focus away from “fault determination” toward identifying the party exercising actual control over risk. The deployer, as the entity responsible for introducing the AI system into commercial, social, or service environments, and as the party directly benefiting from its operation, is therefore required to bear the associated risks.

From a policy perspective, this approach aligns with strict liability regimes applicable to inherently dangerous activities and product liability frameworks in comparative jurisdictions. The AI Law 2025 also conveys a clear message: AI cannot serve as a “technological shield” enabling enterprises to avoid civil liability.

2. What does the principle of “compensate first, seek indemnification later” mean?

A notable feature of Article 29 of the AI Law 2025 is its framework for allocating liability among participants within the AI value chain. This allocation follows principles:

1. The deployer bears primary liability and must directly compensate the injured party; and
2. After providing compensation, the deployer may seek indemnification from the developer, supplier, or other relevant parties only if such indemnification is expressly provided for under contractual arrangements.

The law does not automatically shift liability “upstream” to developers or suppliers. Instead, risk allocation depends entirely on contractual arrangements among parties within the AI ecosystem.

In practice, particularly in multinational enterprise structures, the AI value chain may include:

1. An R&D center in Vietnam responsible for developing AI models;
2. A parent company or affiliated entity responsible for commercial deployment; and
3. Third party vendors involved in training, supplying, or integrating AI systems.

Absent clear contractual provisions governing liability allocation, limitation of liability, and indemnification obligations, the deployer in Vietnam may become the ultimate risk bearing entity.

3. Why are liability exemptions under the AI Law 2025 not a reliable legal shield?

The AI Law 2025 recognizes only 02 limited grounds for exemption from liability:

1. The damage arises entirely from the intentional fault of the injured party; or
2. The damage results from force majeure or a situation of necessity.

Notably, these exemptions are narrowly defined. In practice, proving that the damage was caused “entirely by the intentional fault” of the user is not straightforward, as the boundary between technical failures, model errors, and force majeure events remains subject to significant debate.

Accordingly, in practice, the statutory exemptions do not function as a “legal shield” for enterprises, but rather serve as limited exceptions. The burden of proof largely remains with deployers and other relevant parties.

4. How do regulatory inspections affect AI liability exposure?

Article 28 of the AI Law 2025 grants Government authorities the power to conduct in-depth inspections of AI systems, including the authority to require disclosure of:

1. Technical documentation;
2. System logs and traceability records;
3. Training datasets; and
4. Other relevant information necessary to determine the cause of incidents and allocate liability.

This provisions demonstrate that inspection and examination are not only as administrative enforcement mechanisms but also as a foundation for determining civil liability for compensation in civil disputes related to AI. For enterprises, particularly technology companies:

1. Lack of traceability mechanisms within AI systems;
2. Incomplete technical documentation; or
3. Inability to demonstrate effective risk management and control procedures.

In such cases, enterprises may face substantial liability exposure even in the absence of clear legal violations. Moreover, the inability to demonstrate adequate control and explainability may give rise to adverse inferences regarding liability.

Legal Support from BLawyers Vietnam

AI presents not only technical risks but also legal and financial risks. If your enterprise is researching, developing, deploying, or operating AI systems in Vietnam from 2026 onward, BLawyers Vietnam can assist in proactively managing AI related legal risks, including:

1. Advising on identifying the liable parties for compensation within the AI value chain;
2. Advising on and developing mechanisms for risk allocation and compensation liability in AI related contracts;
3. Providing legal advice on AI risk governance, incident handling, and accountability obligations in cases involving AI-related damages;
4. Assisting with the review of technical documentation, operational logs, and AI monitoring frameworks; and
5. Representing and supporting enterprises in engagements with Government authorities, business partners, and affected parties in connection with AI related incidents.

The above is not official advice from BLawyers Vietnam. If you have any questions or suggestions about the above, please contact us at info@blawyersvn.com. BLawyers Vietnam would love to hear from you.

Date: 24 March 2026

Writer: Ha Le