Resources
Notes on controlling what AI agents do.
Short, practical guides for engineering, business, risk, and compliance teams. Educational only — not legal advice.


Latest
AI agent security: a practical guide for teams shipping agents
AI agent security is about controlling what agents can do, not just what they say. The risks, the controls that work, and where to start.
Security5 min read
Read articleMore articles

AI agent use cases: 8 common agents and the limits each one needs
The most common AI agent use cases in business — support, finance, sales, IT, operations — and the specific limits each needs before going live.
Use cases3 min read

AI agent policy: how to write rules your agents actually follow
An AI agent policy sets what agents may do alone, with approval, or never. How to write one, who owns it, and how to enforce it on every action.
Agent policy4 min read

Prompt injection in AI agents: how it works and how to contain it
Prompt injection turns AI agents against their owners. How direct and indirect attacks work, and why the best defence sits at the action layer.
Security4 min read

Deploy AI agents to production: a readiness checklist
Before you deploy AI agents to production, check identity, limits, approvals, testing, monitoring and rollback. A practical readiness checklist.
Engineering4 min read

How to test AI agent guardrails before production
Test AI agent guardrails by replaying real agent activity through a new rule before it goes live, so limit changes never surprise anyone.
Engineering2 min read

AI agent approval workflow: routing, context, timeouts and escalation
A good AI agent approval workflow gets the right person a clear request fast, and fails safe. How to design routing, context, timeouts and escalation.
Approvals3 min read

LLM guardrails: content filters vs. action controls
LLM guardrails filter what models read and write. Here's what they're good at, what they miss once agents take actions, and how to cover the gap.
Security3 min read

AI agent compliance: the controls and records reviewers ask for
AI agent compliance comes down to provable controls — enforced limits, human oversight and complete records. What reviewers ask for and how to be ready.
Compliance3 min read

AI agent permissions: deciding what an agent is allowed to do
AI agent permissions are more than API scopes. How to decide what each agent may do alone, with approval, or never — and enforce it per action.
Security3 min read

The EU AI Act and AI agents: human oversight, logging and deadlines
How the EU AI Act relates to AI agents — risk categories, human oversight, logging and the 2027–2028 deadlines. A plain-language overview, not legal advice.
Compliance3 min read

Least privilege for AI agents: a practical checklist
Least privilege for AI agents means narrow tools, their own identity and per-action limits. A practical checklist to shrink an agent's blast radius.
Security3 min read

AI agent audit trail: what to record for every action
An AI agent audit trail should answer one question in minutes — what happened, and who allowed it? Here's exactly what to record.
Audit3 min read

MCP security: protecting agents that use Model Context Protocol tools
MCP security covers tool poisoning, rug pulls, confused deputies and token misuse. What the risks are and how to control MCP tool calls in practice.
Security3 min read

Policy as code for AI agents: versioned rules you can review and roll back
Policy as code for AI agents moves limits out of prompts and tool wrappers into versioned rules you can review, test and roll back like software.
Agent policy3 min read

AI customer service agent refunds: letting support agents refund safely
AI customer service agent refunds cut wait times — if limits are enforced. How to set refund thresholds, approvals and fraud checks that hold.
Use cases3 min read

LLM function calling security: validating tool calls before they run
LLM function calling security starts with one rule — never trust a tool call because the model made it. How to validate arguments, scope and intent.
Security3 min read

AI agent governance: a practical operating model
AI agent governance turns principles into daily practice — owners, rules, approvals and records for every agent. A lightweight model you can run.
Agent policy3 min read

LangGraph human in the loop: interrupts, approvals and where policy fits
LangGraph human in the loop uses interrupts to pause a graph and resume it with a person's decision. How it works, common pitfalls, and where policy fits.
Engineering3 min read

Human in the loop for AI agents: when to ask a person
Human in the loop works for AI agents when it's reserved for the actions that matter, routed to a named approver, and fast to answer.
Approvals3 min read

Agentic AI risks: what can go wrong when AI starts acting
Agentic AI risks go beyond bad answers — goal hijack, tool misuse, privilege abuse and untraceable actions. What they look like and how to reduce them.
Security3 min read

AI agent payments: controls for agents that move money
AI agent payments are arriving fast. The controls companies need when agents pay invoices or buy on their behalf — limits, approvals and records.
Use cases3 min read

AI agent risk assessment: a template for reviewing agents before launch
An AI agent risk assessment looks at what an agent can do, not just what it says. A practical template to review agents before launch.
Compliance3 min read

OpenAI Agents SDK guardrails: what they cover and where to add policy
OpenAI Agents SDK guardrails validate input, output and tool calls, with approvals for sensitive tools. How they work and where centralised policy helps.
Engineering3 min read

Excessive agency: the OWASP LLM risk behind most agent incidents
Excessive agency is when an agent has more functionality, permissions or autonomy than it needs. How OWASP defines it and how to fix it.
Security3 min read

ISO 42001 for AI agents: mapping runtime controls to an AI management system
ISO 42001 sets up an AI management system. How AI agents fit in, which Annex A areas matter most, and the evidence runtime controls give auditors.
Compliance3 min read

AI agent guardrails: why prompts aren't limits
AI agent guardrails written in a prompt can be talked past. Here's how to enforce real limits on what agents do, checked on every action.
Agent policy3 min read

AI agent data access: limiting what agents can read and share
AI agent data access controls stop agents reading or sharing more than they should. Patterns for scoping lookups, masking personal data and blocking exports.
Use cases3 min read

Human on the loop vs human in the loop: choosing the right oversight for AI agents
Human on the loop vs human in the loop: one approves actions before they run, the other monitors and can intervene. When to use each for AI agents.
Approvals3 min read

AI agent observability vs. audit: what to monitor and what to record
AI agent observability helps you debug and improve agents; an audit trail proves what they were allowed to do. What each covers and why you need both.
Audit3 min read