Business Pill · The smallest key that works
Least privilege means giving each AI agent only the access its task needs, and nothing more. When an agent is tricked, the damage is as large as its key, so the key should be small.
A short explainer video, under a minute. The intern, the attacker and the robot are an illustration.
At a glance
What Least Privilege Is
The rule that prevents this is called least privilege: give each agent only the access its task needs, and nothing more.
What It Looks Like in Practice
The video's board lists three things. The agent gets its own identity, not a borrowed one.
Why the Size of the Key Matters
The reason is what happens when an agent is tricked.
The Short Answer
Picture an intern who is handed the chief executive’s master key on the first day, to save time. Nobody would do that.
Yet companies do it with AI agents every day. The agent needs to book a meeting room, and it is given the manager’s full login.
What Least Privilege Is
The rule that prevents this is called least privilege: give each agent only the access its task needs, and nothing more.
A meeting room needs a calendar. It does not need a manager’s whole account.
What It Looks Like in Practice
The video’s board lists three things. The agent gets its own identity, not a borrowed one. Its access is narrow. And every action is logged under its own name.
Each one makes the agent’s reach smaller, and makes what it did easy to trace.
The Idea in Parts
The lesson in its own parts
What Least Privilege Is
The rule that prevents this is called least privilege: give each agent only the access its task needs, and nothing more.
What It Looks Like in Practice
The video's board lists three things. The agent gets its own identity, not a borrowed one.
Why the Size of the Key Matters
The reason is what happens when an agent is tricked.
Why the Size of the Key Matters
The reason is what happens when an agent is tricked. In the video, an attacker tries to talk the agent into handing something over.
When an agent is tricked, the damage is as large as its key.
Why It Matters
The video turns this into one question: what can our agent reach that its task does not need?
An agent can do whatever its access allows. So give it the smallest key that works.
The key insight: An agent can do whatever its access allows. So give it the smallest key that works.
Where This Shows Up in AI Business Today
From our news coverage
OpenAI Says It Disrupted a Distillation Campaign
Access that looks ordinary can still be the exposure. In our report, OpenAI says the operators of a coordinated campaign did not break its encryption, compromise a database or gain direct access to stored user conversations.
From our news coverage
OpenAI says they instead manipulated model interactions so that protected reasoning could be reproduced in forms visible to the requester, in a coordinated, scaled way that violated its terms of service.
The Business Engineer Lens
Where this idea sits in the Business Engineer library:
The Question to Ask
- What can our agent reach that its task does not need?
How This Concept Connects
Layer 5 of 8 · Trust, Risk & Governance, on the Business Engineer Concept Map
- Least Privilege governs AI Agents. An agent can do whatever its access allows, so least privilege gives it only the access its task needs.
- Least Privilege governs Computer Use. If a person can click it, an agent can too, so access must be narrowed to the screens the agent may touch.
- Least Privilege constrains Prompt Injection. When an agent is tricked, the damage is bounded by its access, so the smallest key limits what injection can do.
- Least Privilege builds on Declared Identity (lesson unlocks soon). Least privilege gives each agent its own identity and logs every action under its own name.
- Least Privilege constrains Attack Surface (lesson unlocks soon). Giving each agent only the access its task needs closes doors it does not need, the attack-surface advice.
Inside Anthropic’s Permission Layer
Access gating is this framework’s subject.
Get the full Concept Map as a free poster.
Related Business Pills
All Business Pills
- The Memory Wall: Why a Faster AI Chip Is Not Faster AI
- Tokens per Watt: What an AI Data Centre Actually Produces
- Why AI Can’t Be Both Instant and Cheap: Latency vs Throughput
- The Model and the Harness: Why Same-Model Products Differ
- Context, Not Capability: Why a Smart AI Model Gives Poor Answers
- Discardable Software: When Code Is Cheap Enough to Throw Away
- Human in the Loop vs Human on the Loop: Supervising AI Agents
- The AI Audit Problem: When Making Work Is Cheaper Than Checking It
- AI Evaluation as Acceptance Test: How to Know It’s Good Enough
- Extensibility Is Control: Who Holds the Power in an AI Product
- RLHF Explained: How an AI Model Learns What People Prefer
- Pretraining Explained: How an AI Model Learns Before Anyone Teaches It
- Fine-Tuning Explained: How to Adapt a General AI Model to One Job
- Tokens Explained: The Unit AI Reads, Writes and Bills In
- Embeddings Explained: How a Machine Compares Meaning
- RAG Explained: How a Model Answers From Your Documents
- AI Hallucination Explained: Why Models State False Things
- Distillation Explained: How a Small Model Learns From a Large One
- Reasoning Models Explained: What Changes When AI Thinks First
- Tool Use Explained: How an AI Model Goes From Text to Action
- Capex and Depreciation Explained: Why Chip Lifetime Drives AI Profits
- Run-Rate Revenue Explained: What an AI Company’s Number Means
- Backlog Explained: Revenue That Is Signed but Not Yet Earned
- Switching Costs Explained: Why It Is Hard to Leave an AI Supplier
- Temperature Explained: The Dial That Sets How Predictable an AI Model Is
- Guardrails Explained: Rules Enforced Around an AI Model, Not Inside It
- Benchmarks Explained: Why a Test Score Is Not Your Own Result
- Agent Memory Explained: How an AI Assistant Remembers You Between Chats
- MCP Explained: The Shared Plug That Connects AI Models to Tools
- Quantization Explained: How a Large AI Model Is Shrunk to Fit
- Parameters Explained: Where an AI Model’s Knowledge Lives
- Scaling Laws Explained: Why AI Labs Keep Building Bigger
- Synthetic Data Explained: When a Model Writes Its Own Training Examples
- Transformers and Attention Explained: How AI Reads Every Word at Once









