AI Dictionary › AI Fundamentals
Rischio Duplice Uso
Dual-use risk describes the characteristic of many AI capabilities to serve both legitimate, beneficial purposes and harmful ones, often with the exact same technical functionality. A model able to write code useful to a developer is also, in principle, able to write malicious code; a system that explains molecular biology for educational purposes shares the same knowledge base that, in the wrong hands, could assist in creating dangerous substances.
Unlike a specific security flaw, dual-use risk is not something that can be fully "fixed" without also giving up the legitimate use of the same capability: the technical challenge is calibrating access, context, and controls to enable the first purpose while reducing the likelihood of the second, rather than eliminating the capability itself.
In modern AI the concept guides decisions about releasing models with advanced capabilities in sensitive areas such as chemistry, biology, cybersecurity, and large-scale persuasion. Labs developing frontier models often publish specific dual-use risk assessment policies before making a new capability public, and some releases are delayed or restricted precisely for this reason.
The term "dual-use" predates modern AI by far, originating in export controls for technologies with both civilian and military applications; it was picked up and adapted by the AI safety community starting in the second half of the 2010s to describe a structurally similar problem in the field of generative models.
From our network
Kaimaki Web: Websites That Win Customers
Custom websites, web apps and digital marketing for growing businesses.
Visit kaimakiweb.com →From the Agora Intelligence blog
📱 Download the Android app (beta) iOS coming soon
Say what you mean. Get what you need.
Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.