AI Dictionary › AI Fundamentals

Least Privilege (AI)

Principio del Privilegio Minimo (AI)

The principle of least privilege applied to AI states that an agent or application built on a language model should have access only to the resources, data, and tools strictly necessary to perform its assigned task, and nothing more. It is not a model-specific security technique, but an architectural criterion for designing the system around it.

Definition

How it works

In practice, applying least privilege means giving an AI agent granular, specific permissions rather than generic credentials: an assistant that needs to read a calendar should not be able to delete events; a bot generating reports from a database should have read-only access, not write access; an agent browsing the web for research should not hold the same credentials that would let it make payments. The principle applies both to the agent's direct permissions and to the tools exposed to it via function calling.

Applications

It is one of the most effective defenses against the consequences of a successful attack, such as a prompt injection: if an attacker manages to manipulate the agent but the agent only has minimal permissions, the potential damage stays contained even when the manipulation succeeds. It becomes increasingly central as AI agents capable of autonomous multi-step actions spread, where every excess permission granted is an additional attack surface.

History & etymology

The principle of least privilege originates in classic computer security, explicitly formulated as early as the 1970s for designing secure operating systems; it was brought back to the center of the debate with the spread of autonomous AI agents starting in 2023, when giving them access to real tools became common practice.

Related terms

• AI Agent
• Tool Use
• AI Kill Switch
• Indirect Prompt Injection
• Runtime guardrail

More in AI Fundamentals

• LLM (Large Language Model)
• Token
• Context Window
• Temperature
• AI Hallucination
• Fine-tuning
• RAG (Retrieval Augmented Generation)
• Embedding

Put it into practice

• Real scenario: apply this concept with an instant AI verdict.
• Daily gym: short drills to train dictionary concepts.
• Playoff game: challenge other players, elimination-style.

📱 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.