The frontier model companies are responsible for how their products perform in the real world, and their customers are responsible for how they use them. If a user of Claude runs a cyberattack — especially one at scale — both the customer and the platform will be sued. This should be enough of an incentive to force these companies (let’s stop calling them labs) to produce safe products AND monitor for unsafe use (i.e. ie hacking). These frontier models have the option to lock them down……