Artificial Intelligence
Trending
What Are AI Guardrails and How Do You Add Them?
What AI guardrails are, the 3 layers every AI app needs (input filters, output checks, permission limits), and a step-by-step plan to add them.
Cybersecurity
What is Jailbreak?
Quick answer
Bypassing the restrictions on a device or an AI model, either to remove manufacturer limits or to make a model ignore its safety rules.
Jailbreak, explained
Originally, jailbreaking meant removing software restrictions on devices like smartphones so users could install unapproved apps or customize the system. This can void warranties and weaken security, since it disables protections built into the operating system.
In the context of AI, a jailbreak is a carefully crafted prompt designed to trick a large language model into ignoring its guidelines, for example by role-play scenarios or confusing instructions. AI developers treat jailbreaks as a security and safety issue and continually test and update models to resist them. A related threat is prompt injection, where malicious instructions are hidden in content an AI agent reads, such as a web page or document, to hijack its behavior.
Related terms
AI Agent
A software system that uses an AI model to plan and carry out multi-step tasks, often by calling tools, APIs, or other programs on its own.
AI GuardrailsRules, filters, and checks placed around an AI system to keep its inputs, outputs, and actions within safe and acceptable limits.
Prompt InjectionAn attack that hides instructions in text an AI system reads, tricking it into ignoring its original rules or taking unintended actions.
Guides that mention Jailbreak