TLDR; Real-world crises often fail because of poor coordination between key players. Tabletop exercises (TTXs) are a great way to practice for these scenarios, but they are typically expensive and exclusive to elite groups.

To fix this, we built a free, AI-powered, single-player digital TTX that simulates an election misinformation crisis. It's a "rehearsal space" designed to make high-stakes decision-making and systems thinking accessible to everyone. We're inviting you to try the prototype, give feedback, and share your thoughts on TTX in general. 

Why bother?

In the summer of 2025, as the world seemed to grapple weekly with new LLM releases and ever-expanding AI deployments, a small group of us gathered in India. We met not to keep pace with the frenzy, but to ask a different question: how do we educate for the future of AI? In the middle of so much rapid change, are we seeing light at the end of the tunnel—or the headlamp of an oncoming train?

The room was diverse: some of us came from technical backgrounds, others from philosophy or policy. Each brought a different lens, but we shared the same unease. AI’s ubiquity has made it central to diplomacy, trade disputes, national security, and questions of sovereignty. And if we wanted to think seriously about intervening—or even just understanding—such a systemic force, we needed tools for exploring not only individual actions but also the consequences that ripple outward across entire systems.

One of us had previously taken part in tabletop exercises (TTXs): a biosecurity drill for pandemic preparedness and a foreign policy scenario on AI deployments. Drawing from that experience, and impatient with writing cards and rules by hand, we built a digital version—a simulation of an election crisis. This experiment gave us a glimpse of how TTXs could be adapted for broader communities, helping participants see how risks play out in complex, sometimes catastrophic ways, while also expanding the space of possible scenarios.

The reason this mattered became clear quickly. In real-world crises—whether climate shocks, bio-risk, cyberattacks, or AI misalignment—one failure mode repeats itself: poor coordination between key actors under uncertainty. Governments, corporations, researchers, and civil society often work in silos, with fragmented incentives and little practice at real-time collaboration as exhibited during the coronavirus pandemic in 2020.

Tabletop exercises, at their core, ask participants to role-play as stakeholders in a scenario under time pressure. Each decision carries consequences, often nonlinear, and together they generate emergent effects. Militaries have long used them as war games; industries and policymakers have used them to sharpen strategy and coordination, also business schools train graduates in simulated environments to sharpen strategy skills. 

Those who’ve participated in TTXs—whether in the military, industry, or disaster response—consistently find the same thing: when people rehearse crises, they gain sharper teamwork, clearer roles, and more confidence under pressure.But the catch is they’re usually conducted in person, require hours, and are logistically demanding—practical mostly for elite groups such as army officers, senior management and policymakers.

Where We Stand Now

So far, we’ve built one scenario: a single-player simulation of misinformation during an election crisis, powered by AI but scaffolded by us. 

In this AI-powered simulation, you'll choose a role and face an escalating scenario. You must make tough choices with limited resources to advance your secret objectives while maintaining public trust. An AI Game Master generates the story, controls the other characters, and shapes the consequences of your actions, ensuring a unique challenge every time. Your goal is to navigate the crisis and learn about high-stakes, multi-stakeholder decision-making.

Try the game - https://ai-risk-ttx.vercel.app 

Or if you want to contribute, send a PR to https://github.com/bhi5hmaraj/ai-risk-ttx 

What This Simulation Helps Us Practice

This simulation is designed as a rehearsal space where participants can experience three kinds of learning at once. 

1. It makes systems thinking tangible, as actions reveal hidden frictions, cascading risks, and leverage points that would otherwise remain abstract.
2. It also cultivates reflex under uncertainty, pressing participants to exercise judgment in time-sensitive and ethically complex situations—developing the kind of muscle memory that written strategies alone cannot provide. And because the format is accessible,
3. it democratizes foresight: moving a tool once reserved for militaries and policy elites into classrooms, nonprofits, startups, and think tanks.

At its heart, this work begins with a simple recognition: we are not yet prepared for the coordination challenges that define this century. It is also difficult to fully grasp the perspectives of different stakeholders when discussing shared risks or common goals. Tabletop exercises create that opportunity, allowing participants to step into the roles of both adversaries and allies, uncovering blind spots and experimenting with strategies for better coordination. While these exercises have long been trusted as a way to build foresight, they remain rare and confined. To meet the scale of today’s risks, they must become more open, adaptable, and widely shared. This experiment is one step in that direction—combining the craft of simulation design with digital accessibility to create not just a game, but a rehearsal space for tracing the hidden machinery of society—the gears that grind, catch, and sometimes misfire when crises force us to move together.

The current prototype offers a positive step toward practicing decisions in complex, high-risk settings. Still, it raises questions. What about the “unknown unknowns”—the risks that can’t be simulated, only surfaced in conversation?  Could relying on large language models to build simulations introduce its own vulnerabilities? With only a single player, can we meaningfully capture the emergent dynamics that make these exercises powerful? And is it responsible to let AI shoulder the work of scoring? Perhaps the answer is a hybrid model, combining digital tools with in-person exercises - and adding layers of thoughtful and research backed game design.

For now, we’d like to invite you in. Try the prototype. Stress-test it. Tell us what it reveals—and where it falls short. We don’t know exactly what rehearsing the future should look like, but we do know this: it’s better to stumble in simulation than in reality.

If TTX and systems thinking within AI governance interests you then look up AI2027,  AI Takeoff Report by Tom Davidson, AI Futures project, Intelligence rising.