✳flâneur — a map of the web's best reading

AI Policy Bulletin

aipolicybulletin.org · 130 words · saved by 1 readers

Current AI safety frameworks evaluate for certain 'canonical' dangerous capabilities – like chemical, biological, radiological and nuclear (CBRN) weapons, cyber operations, and misinformation. Despite growing efforts to evaluate frontier AI models for these known dangerous capabilities, we may be missing entire classes of unknown threats. Legal Zero-Days are one previously overlooked category – a novel threat vector through which advanced AI systems, either through misuse or loss of control, might bypass safeguards or accumulate power by exploiting unforeseen vulnerabilities in complex legal systems. An example helps illustrate the idea. Since 1901, the Australian Constitution has prohibited parliamentarians from having an “allegiance to a foreign power”. Complex citizenship laws have meant that a place of birth or a relative could confer a citizenship right on an Australian, and hence an allegiance to a foreign power. In July 2017, a Perth barrister highlighted this conflict and provi

AI Policy Bulletin Log In Lorem ipsum dolor sit amet. Log in with Google Log in with Facebook Log in with Apple Thank you! Your submission has been received! Oops! Something went wrong while submitting the form. Forgot your password? Don't have an account? Sign Up Sign Up Lorem ipsum dolor sit amet. Sign up with Google Sign up with Facebook Sign up with Apple Thank you! Your submission has been received! Oops! Something went wrong while submitting the form. Already have an account? Log In Subscribe Join our newsletter to stay up to date on the latest AI policy articles we publish! Have somethi

Explore this link on the map →

related reading