Malaysia Wants Safety by Design for AI Chatbots Aimed at Children
Malaysia’s planned AI law will single out chatbots and apps built for children. Developers of AI chatbots and applications aimed at children will have to build in safety features from the start, including regular risk assessments and filters against harmful content, under the proposed Artificial Intelligence Governance Bill, the Digital Ministry said in a written parliamentary reply reported by The Star on October 6.
“For applications or chatbots that target children, developers must apply safety-by-design elements,” the ministry said. That includes making periodic risk impact assessments mandatory and filtering harmful content to prevent the psychological manipulation of children. The reply answered a question from Riduan Rubin, the MP for Tenom, who asked how the Bill would protect children from chatbot risks and whether bot disclosure would be made compulsory to protect minors.
On disclosure the answer was more measured. The ministry said requirements would be imposed in a targeted way, depending on an application’s risk level and the needs of each sector. AI service providers will eventually have to display clear notices that users are interacting with AI, especially for applications classified as high risk. Details could be set through sector-specific laws or codes of ethics issued by a regulator that the Bill would create.
What it means in practice
The approach is risk-based, so a homework helper and a companion chatbot may not face identical rules, and much will depend on how the regulator classifies children’s products. The ministry said it consulted the Women, Family and Community Development Ministry and Unicef in developing the Bill. Digital Minister Gobind Singh Deo said last month that the draft was complete and that he hoped to table it in Parliament by the first quarter of next year at the latest, a timeline AI4Malaysia followed when the draft AI Governance Bill was completed and when Gobind said it was headed to Cabinet.
For developers, safety by design is an engineering brief as much as a legal one. Age-appropriate defaults, content filtering, documented risk assessments and clear AI labels are easier to build in early than to retrofit after launch, and companies serving Malaysian families with tutoring, gaming or chat products have time to review their systems before the Bill is tabled. Concern about AI and manipulation is already on the regulatory agenda here: the communications regulator recently opened a probe into a report on AI-driven voter manipulation.
What to watch: the final wording of the children’s provisions when the Bill is tabled, which body becomes the AI regulator, and whether high-risk categories are defined in the Bill itself or left to later codes.