Technology governance scholar Gillian Hadfield on AI safety and security

In the excerpt, Hadfield discusses her work on the legal, regulatory and technical infrastructure for AI systems that operate within human normative frameworks. The excerpt summarises her remarks rather than presenting new empirical findings.

The excerpt describes one component of that work: a "regulatory markets" idea. Under the model as described, governments establish what is acceptable risk, and then license private actors who show they have developed the tests and verification to demonstrate that a system is safe. The arrangement divides responsibility: public authorities define the threshold of acceptable risk, while private organisations supply the testing and verification capability that demonstrates compliance with it. Hadfield is described in the excerpt as working on the legal, regulatory and technical infrastructure for AI, with safety and security framed as governance questions.

The excerpt provided does not name specific jurisdictions, legislative instruments, agencies, timelines or companies, and does not report quantitative findings; the account of the "regulatory markets" idea is given at the level of principle. The framing situates AI safety and security within governance design — the rules, institutions and verification arrangements surrounding a system — rather than within model development or technical alignment research alone. Beyond the licensing model described, the excerpt does not specify how acceptable risk would be defined, who would conduct the licensing, or what standards the tests and verification would be required to meet.