Democratic scrutiny is public or institutional oversight of consequential decisions, supported by transparency and accountability. In technology and cybersecurity, it helps people assess who made a decision, what evidence informed it, and who is responsible for its consequences.
For advanced AI, scrutiny can include publishing clear system definitions, documenting risk assessments and pre-deployment tests, and identifying review and incident-response responsibilities. These measures do not prove a system is safe, but they make safety claims more verifiable and enable regulators, public institutions, and affected communities to question them. Without accessible evidence and accountable decision-makers, oversight risks becoming a matter of trust rather than a check on technical and institutional power.