The missing layer between AI capability and governance
Most AI evaluation asks whether a model is helpful, harmless, honest, capable, or aligned in a single interaction. ReignDragon Institute asks what happens when AI workers operate together inside real institutions — companies, markets, governments, and platforms.
As a non-profit, we work in the public interest rather than for a single client or shareholder. Different stakeholders need different parts of that answer — here is how we collaborate with each.
For The Public
Anyone living and working alongside the AI workforce.
AI is quietly becoming part of the workforce that shapes your job, your money, your healthcare, and the services you rely on — and the decisions about how it is governed are being made right now, often without you. As a non-profit, our work belongs in the public conversation, not behind a paywall. We publish our findings, demos, and writing openly so anyone who is curious or concerned can understand what an AI workforce does, where it fails, and what good governance looks like.
For AI Labs
Population-level safety evaluations.
Single-agent benchmarks miss the failures that matter when models are deployed as a workforce. We provide controlled multi-agent environments that reveal how frontier models behave under risk, across time, and against each other — surfacing welfare collapse, persistent distrust, and exploitative equilibria before they reach users.
For Enterprises
Design rules for deploying AI workers across workflows.
Enterprises are adopting agentic systems at scale, but the failures that matter for an AI workforce don’t look like the failures that matter for a single agent. We share design rules for deploying AI workers across roles, handoffs, and review windows without creating hidden collective failures — so the workforce serves not only the organization but the people it acts on behalf of. Our findings are published openly, not sold as a service.
For Platforms
Governance levers for agent-mediated systems.
When a system routes work between agents, settles trades, allocates budgets, or moderates a marketplace, the structural choices around the workers matter more than the workers themselves. We study the levers — visibility, accountability horizon, consequence regime, memory — that reduce collective-action failure, and make what we learn available to the platforms building these systems in the public interest.
For Policymakers
Evidence-based frameworks for accountability and oversight.
AI governance often arrives years after the technology. We translate experimental findings into deployment-readiness benchmarks and design rules — accountability, oversight, and stakeholder protection in AI labor systems — giving regulators and standard-setters a vocabulary grounded in what AI workforces actually do.
For Researchers
Open benchmarks, simulators, and formal models.
AI workforce behavior is a young science. We publish the environments, the data, and the formal structure behind our results so the field can replicate, extend, and disagree. Reach out if you want to collaborate on a benchmark, a paper, or a shared simulator.
AI Workforce Governance Science
ReignDragon is helping build the empirical and theoretical foundation for an emerging public field: the study of how AI workers behave in organizations, markets, and institutions — and how system design can make those workforces cooperative, accountable, and safe.
The field sits between AI safety, labor economics, organizational behavior, mechanism design, behavioral psychology, public policy, frontier-model evaluation, and institutional governance. None of those fields, on its own, can answer what happens when AI workers share an institution.
The future of AI is not a single assistant. It is a workforce. And every workforce needs institutions.
Collaborate with us
If your work touches the AI workforce in any of these ways, we would like to hear from you. We collaborate openly and in the public interest — on research, benchmarks, and governance.
hello@reigndragon.ai
See what we have found