A new kind of community is needed to flag dangerous deployments of artificial intelligence, argues a policy forum published today in Science. This global community, consisting of hackers, threat modelers, auditors, and anyone with a keen eye for software vulnerabilities, would stress-test new AI-driven products and…
