AppliedAI360
Labs

Labs

Working models, not slides. Each lab shows how an advisory approach behaves on realistic systems — click through it, change things, and see the consequences.

LiveAdvisory: Agentic & runtime governance →

Can this AI agent be trusted?

An AI agent that pays supplier invoices is tricked by a fake invoice. Watch the checks stop the payment, the agent’s authority drop automatically, and the bank refuse to let it act alone — the full loop of verifying an AI agent before and while it works.

What you can explore
  • ID card: who the agent is, who is responsible for it, and what it is built from
  • Job & limits: exactly what it may do, and how much it may spend
  • Entrance test: 1,240 situations, including the most common ways agents are tricked
  • Live checks: every action checked before it happens — try tricking it yourself
  • Trust level that drops automatically when the agent misbehaves
  • The bank’s check, and an audit trail nobody can secretly change
Frameworks
OWASP Agentic Top 10NIST AI RMFEU AI ActISO/IEC 42001

Three fictional AI agents at a fictional lender: supplier payments, debt collection and customer checks. The digital signatures and locked records use real cryptography in your browser.

Open the lab →Read the researchDiscuss it for your organisation

Fictional company and synthetic data. Framework references are indicative and not legal advice.

LiveAdvisory: AI governance programme design →

AI Governance Workbench

Every AI application against the EU AI Act, ISO/IEC 42001, NIST AI RMF and the OECD AI Principles — in one matrix. Adherence is decided by runtime evaluation metrics measured on each application, with documents as secondary evidence.

What you can explore
  • Organisation matrix: % adherence for every application × framework
  • Drill down by framework or by application to each control
  • Runtime metrics and documents side by side, with evidence for every finding
  • Industry-standard metrics: OWASP LLM Top 10, RAGAS, DeepEval, C2PA, Fairlearn and more
  • Fix once, close many — remediation ranked across all frameworks
  • What breaks if…? — preview the governance impact of a model, prompt or tool change
Frameworks
EU AI ActISO/IEC 42001NIST AI RMFOECD AI Principles

Four GenAI applications: a RAG assistant, a high-risk letter writer, an agent with tools and a generative-media studio.

Open the lab →Discuss it for your organisation

Fictional company and synthetic data. Framework references are indicative and not legal advice.

In the lab next