National Assembly Debates Who Is Liable When AI Gets Electronic Identity Checks Wrong

Vietnam’s National Assembly debated the draft Law on Electronic Identification and Authentication in plenary session on the morning of November 4, and one of the bill’s new elements is a provision on using artificial intelligence in identity checks, VTV reported. AI can speed up verification for opening bank accounts, making online transactions and using digital services, but lawmakers and experts say the rules need to be clear about accuracy, safety and the rights of users.

Clause 4 of Article 43 of the draft says the unit that manages an AI system is responsible for its technical accuracy, while the unit that uses the identification result bears final responsibility. Some experts have asked who would be liable if a wrong AI verification led to someone losing money, having an account locked or having their identity stolen, especially when the organisation relying on the result neither built nor controls the algorithm. Lawyer Nguyen Ngoc Hung, head of the Ket Noi law office, said a wrong result can stem from an inaccurate algorithm, incomplete input data, a faulty recognition system or a provider’s poor operating procedures.

Others defend the split. Nguyen Binh Minh, of the senior advisory council of the Vietnam E-commerce Association, compared it to buying a tool: if an accident happens because it was used wrongly, the user is responsible, not the manufacturer, unless the fault lies in the product itself. Journalist Nguyen Tat Hong Duong, editor-in-chief of the Vietnam Cybersecurity electronic magazine, said it is getting harder online to tell what is real, as accounts impersonate acquaintances, websites imitate banks and voices can be faked with AI, which makes trustworthy authentication more important.

What it means in practice

The debate matters because AI is spreading quickly through the public sector. At its monthly press conference the same day, the Ministry of Science and Technology said its AI platform supporting public duties is in wide trial at all 24 ministries and central agencies and all 34 provinces and cities, with more than 539,000 accounts issued, VTV reported. Liability questions also run through the draft data security law’s rules on AI training data and the State Bank’s risk tiers and human review rules for banking AI.

What to watch is whether the final text spells out how citizens can claim compensation after a faulty AI check, and how the split of responsibility will work for banks and service providers that buy AI verification from third parties.