Intelligence-grade AI-risk exercises

Rehearse the hard AI decisions before they cost you

Furioso puts your leadership team inside a realistic AI-failure scenario and surfaces, on the record, where your decisions break down while it is still cheap to fix them.

What it is

A facilitated exercise, not a workshop

An AI-risk tabletop exercise is a structured session that tests how your team would actually respond when an AI system fails, misfires, or gets trusted when it shouldn't be. The facilitator runs realistic injects and actively probes the team's reasoning rather than leading a discussion.

You leave with a clear picture of where authority, escalation, and communication break down, and a written record you can act on.

How the method works
  • Surfaced decision gaps. The points where your response chain stalls or breaks, made visible on the record.
  • A rehearsed response. Your team has now run the scenario once, in a safe room, before it happens for real.
  • A written after-action report. Findings, owned recommendations, and the early-warning indicators to watch.

Who it is for

Built for the teams that own the decision

Three scenarios drawn from the risks leading 2026 agendas.

Enterprises deploying AI

Agentic systems and shadow AI in financial and operational workflows. Test detection, escalation, kill-switch authority, customer comms, and regulator notification.

Government & policy

A decision-support model trusted for a high-stakes public choice turns out to be biased or manipulated. Test accountability and how the institution reconstructs what it relied on.

Boards & executives

A 90-minute pilot that gives time-poor directors a fast, concrete read on whether the organisation is ready for the AI risks it has taken on.

EU AI Act Article 4 enforcement starts 2 August 2026. If your staff use AI, the literacy duty applies to you.

Literacy workshop Free governance resources

The method

Structured analytic techniques, not improvisation

Each exercise follows the same four steps, drawn from intelligence practice.

1

Pre-work

A scoping call and questionnaire so the scenario fits your sector and situation.

2

Facilitated session

Your team works a realistic AI-failure scenario through timed decision rounds.

3

Decision stress-test

A Key Assumptions Check and scenario injects push the reasoning until it bends.

4

After-action report

Findings, owned recommendations, and indicators to watch — in writing.

Why Furioso

Threat assessment under pressure, now focused on AI

Furioso is led by Erik Bernath: an MA in Intelligence and Security Studies (Brunel), years of threat assessment, and current AI-safety credentials. The same discipline used to assess real-world threats builds realistic AI scenarios and holds a room of executives to them.

More about Erik
MA, Intelligence & Security Studies (Brunel) Intelligence Rising wargame BlueDot AI safety Anthropic Center for AI Safety Author, Minds We Create (2026) Featured in AP News National Law Review

Find the decision gaps while they are still cheap to fix

A 25-minute call to talk through your AI risk and whether a tabletop exercise fits. No pitch deck.