READ · Assurance
Assurance
Independent assessment of AI systems before and after they are released, so organisations and the public can trust what is deployed.
Assurance gives confidence that an AI system does what it claims and behaves within acceptable limits. We assess systems on behalf of industry and public-sector partners across ASEAN, and we publish what we can so that trust is earned in the open rather than asserted behind closed doors.
Focus areas
Pre-deployment evaluation
We evaluate systems before release, testing for capabilities and failure modes that matter for safety, security and societal impact.
Guardrail testing
We test the effectiveness of safety and security features, measuring where they hold and where they can be circumvented.
Red-teaming
We conduct structured red-teaming to find weaknesses before others do, working with developers to close them.
Assurance reports
We publish clear assurance findings that decision-makers can act on, written for non-technical readers as well as specialists.
Instruments
Engage with us
Collaborate with liwaiwai on assurance.
Work with us on research, evaluation or institutional projects. Partner on a study, host a session, or start a conversation.