MPTC Establishes a Cross-Sector Artificial Intelligence Sandbox

In 2025 the Cambodian authorities announced the establishment of a cross-sector artificial intelligence (AI) sandbox, intended to allow controlled testing of innovative AI-enabled products and services in areas including telecommunications, financial services, health, education and public administration. The sandbox is a coordinated initiative between the sector regulator for telecommunications, financial supervisors and other relevant authorities.

The sandbox is relevant to telecommunications operators exploring AI in network operations, customer service and consumer offerings, to fintech providers experimenting with AI-driven credit and payments, to healthcare providers evaluating AI diagnostic tools and to public sector bodies considering AI in service delivery.

The sandbox aims to support responsible innovation by providing a structured environment in which participants can test AI products under conditions that are less restrictive than the fully regulated environment, while ensuring that appropriate safeguards apply for users and third parties. It seeks to accelerate learning about how AI applications behave in real conditions, how they interact with existing regulatory frameworks and where adjustments to those frameworks may be needed.

The sandbox is not a mechanism for exempting participants from applicable laws. It is a mechanism for testing specific arrangements with the knowledge and support of the authorities, so that risks can be identified and managed and so that lessons can inform both individual participants and future policy.

Participation in the sandbox is generally open to organisations that can demonstrate a clear proposal, a well-defined test scope, adequate technical and operational capability and a credible plan for managing the risks associated with the test. Proposals are typically assessed by a joint panel including representatives from the relevant regulators. Applications are expected to describe the AI technology, the intended use case, the target population, the data involved, the safeguards proposed and the outcomes to be measured.

The application process is designed to be collaborative. Applicants are encouraged to engage with the authorities before submitting a formal proposal, to identify the regulatory questions that will need to be addressed and to design the test in a way that is workable for both the participant and the authorities.

Sandbox participants are expected to implement appropriate safeguards, including user consent where required, mechanisms for redress if things go wrong, human oversight of high-impact decisions, monitoring of outcomes and reporting to the authorities. Testing is typically limited in scope, in duration and in the population involved, so that risks are contained and lessons can be learned before any broader deployment.

Data protection is a central consideration. Participants should have a clear legal basis for the data processing involved in the test, should minimise the data used, should apply strong security measures and should ensure that individuals are informed about the test and about their rights.

AI applications frequently cross sectoral boundaries. An AI product that supports credit scoring for telecommunications customers may involve financial supervision as well as telecommunications regulation. A tool that supports health information services may involve health authorities as well as data protection considerations. The sandbox is designed to accommodate such cross-sector applications through joint review and coordinated conditions, reducing the need for participants to navigate multiple sectoral processes independently.

Participants are encouraged to identify the sectoral overlaps in their proposals so that the authorities can allocate responsibility appropriately and so that the test can be designed to address all relevant considerations.

The sandbox is intended to produce learning outcomes that benefit both individual participants and the wider policy environment. Aggregate lessons about the behaviour of AI applications, the effectiveness of safeguards and the interaction with regulatory frameworks are expected to inform future guidance, sector-specific rules and any general AI legislation that may be considered.

Participants are typically expected to share appropriate information about the outcomes of their tests, subject to confidentiality where necessary, and to engage in dialogue about the implications for future practice. The authorities may publish aggregate insights to support broader awareness.

For telecommunications operators, the sandbox opens practical opportunities to test AI applications in areas such as network optimisation, fraud detection, customer support, personalised offers and content management. Operators can use the sandbox to explore the boundary between operational tools that fall clearly within their existing licences and applications that may require additional consideration, particularly where they intersect with financial services, health or other regulated sectors.

Operators should ensure that their internal governance for AI development is aligned with the expectations of the sandbox, that data protection is fully addressed and that outcomes are documented in a way that supports both internal decision-making and engagement with the authorities.

For organisations considering participation, the sandbox is a useful vehicle for exploring innovative uses of AI in a way that manages regulatory risk. Early engagement with the authorities, careful design of the test proposal and readiness to implement meaningful safeguards will improve the likelihood of a productive experience.

For organisations not directly participating, the sandbox is a valuable source of information about how AI applications are being tested in Cambodia, what safeguards are being applied and what lessons are emerging. Monitoring published outcomes can inform internal strategy and risk management.

Lex Civora advises telecommunications operators, financial institutions, health providers and public sector bodies on participation in the AI sandbox, including preparation of proposals, negotiation of conditions with the authorities, design of internal safeguards and translation of sandbox learning into wider strategy and compliance.