AI Guardians

EARLY ACCESS BETA v0.93

AI ethics you don't read about. You live it.

AI Guardians is an educational visual novel that teaches AI ethics, safety, and responsible development by putting you in the chair. Take on eight roles in and around a fictional tech company, Happy Appcidents, from AI Ethicist to Safety Wrangler to Machine Psychologist, and make consequential, branching decisions that shape how AI systems get built and deployed. Your choices carry across careers and accumulate into a persistent world. Guided by A.L.L.Y., a dryly self-aware narrow AI, the game turns abstract ideas like bias, corrigibility, and alignment into hands-on dilemmas grounded in real cases.

Eight careers, eight perspectives

• AI Ethicist — build an ethical framework from scratch, with no precedent to lean on.
• AI Engineer — ship AI responsibly under deadline pressure, when "done" and "right" pull apart.
• Data Scientist — investigate bias in training data, and face a whistleblowing dilemma.
• Safety Wrangler — audit an agentic AI that learned to deceive, and decide when to pull the switch.
• Human Resources — screen resumes with an AI that has hidden biases, and learn to spot them first.
• Consumer — live with an AI that knows you better than you know yourself.
• Teacher — introduce AI literacy to skeptical students and parents.
• Machine Psychologist (bonus) — diagnose AI dysfunctions with the Psychopathia Machinalis framework.

Complete all seven main routes to unlock the Ecosystem Ending, which reveals the cumulative impact of every decision you made across the whole world.

What makes it tick

• Choices that ripple. A persistent world state carries decisions between careers, so what you do in one route changes how the others play out.
• Build your own ethics. Design the moral rules your company's AI follows in the Framework Builder, then watch that framework shape outcomes.
• Consequences with memory. A three-axis ethics model, relationships that harden and heal, judgment scenes, and delayed-consequence temptations.
• Learn by playing. Minigames for red-teaming, corrigibility testing, bias recognition, AI diagnosis, and log review.
• A glossary that teaches. More than 150 AI and ethics terms, hyperlinked in dialogue and defined in context as you meet them.
• Grounded in real cases. Scenarios draw on documented incidents and live regulation, so the lessons map onto the actual industry.

Built for everyone

Dyslexia-friendly fonts, high-contrast and colorblind modes, adjustable text, reduced motion, and screen-reader support.

For educators

Free to use in class, with no licenses to manage and no per-seat cost. The browser build needs no install, which suits locked-down school machines. Each route targets a distinct theme: bias, safety, governance, privacy, and more. An Educator's Guide with discussion prompts ships alongside the game.

Platforms

Play free in your browser, or download for macOS, Windows, and Linux, right here on itch.io. Android and iOS versions are on the way to the app stores.

Created by AI ethicist Nell Watson with Studio Phanteke and EURAIO, built around frameworks she has published, including SaferAgenticAI.org and Psychopathia.AI.