SIRDA

AI safety already has a substantial ecosystem of researchers, evaluators, frontier AI developers, public institutions, nonprofits, funders, educators, and independent thinkers. SIRDA works to understand the strengths within that ecosystem, identify where those strengths can reinforce one another, surface important gaps and opportunities, and inspire action around problems that no single participant is positioned to solve alone.

SIRDA pays particular attention to the convergence of increasingly capable intelligence, AI systems contributing to the development of more capable successors, growing autonomy, access to critical infrastructure, and the ability to act through software and increasingly powerful physical systems. The consequences become especially significant when advanced AI can affect communications, energy, finance, laboratories, industrial systems, cyber infrastructure, robotics, autonomous vehicles, drones, or military systems capable of surveillance, targeting, strategic action, or physical force.

Researchers disagree about how quickly these capabilities will develop and whether they will converge. SIRDA's work holds under any timeline. We study whether humanity has the institutions needed to govern that transition, what capabilities should trigger additional oversight, which powers would be required, who holds those powers today, where important gaps remain, and how those powers could be constrained and coordinated across institutions that may not agree.

The map is designed to show both who is working in the field and what the field can do together. Each Ecosystem page connects people and institutions to their focus areas, selected work, recent activity, stated positions, and documented relationships. Together, those connections can reveal concentrations of expertise, complementary strengths, neglected problems, and opportunities for participants to accomplish more together.

The result is a living research graph showing where knowledge, expertise, and practical resources already exist, how different parts of the field depend on one another, and where new collaborations or institutions may be needed.

SIRDA's editorial account and a participant's own account remain clearly separated. Each participant page can include a SIRDA Overview alongside an Official Profile that the organization or individual can claim and maintain after verification.

The planned Ecosystem Map will visualize these relationships as a constellation of organizations, people, projects, safety frameworks, evaluations, and open questions.

A useful contribution can begin with a research result, source, objection, evaluation, governance mechanism, scenario, or open question. The platform connects those contributions so participants can trace the evidence behind an idea, see the strongest disagreements around it, and understand how one piece of work affects another.

An evaluator's findings may change a governance proposal. A governance dispute may reveal the need for a new technical test. A forecasting result may change the urgency of a safety threshold. A criticism may expose a failure condition that requires redesign. Expertise in one part of the ecosystem may provide the missing piece for a problem somewhere else.

The alliance can help identify where existing institutions are especially well positioned to act, where complementary strengths could be brought together, and where an important problem has no adequate institutional home.

Those gaps also tell us something about governance. They show which powers already exist across governments, companies, infrastructure providers, evaluators, markets, and the public; where those powers could be coordinated more effectively; and where an important function has no holder at all. That work is Stage One of SIRDA's founding inquiry and, if the hypothesis survives criticism, the starting material for constitutional design.

The goal is coordination under disagreement: enough shared structure for people with different assumptions, specialties, and priorities to investigate the same high-consequence problems and discover what they can do together.

The human problem is deciding what to do when governments, companies, evaluators, researchers, and affected communities see the same evidence and reach different conclusions.

Civilization already contains technologies for coordination under disagreement: constitutions, courts, rights, appeals, auditing, scientific review, treaties, verification, and separation of powers. SIRDA is investigating how those principles may need to evolve for artificial intelligence operating at global scale and potentially machine speed.

The hypothesis is a government organized around a specific class of dangerous capabilities rather than a geographic territory. Its mandate would be limited to advanced AI systems capable of imposing catastrophic or existential risk across borders. Existing nations, companies, research institutions, and communities would continue to govern the rest of human life.

The inquiry examines the hard questions first: where legitimate authority could come from, what activates jurisdiction, how disputed evidence becomes a decision, what rights affected parties retain, how power is distributed, how capture and collusion are constrained, how the institution relates to existing states, and what evidence should cause the project to be abandoned.

If the inquiry survives public criticism, the next stage is collaborative constitutional design. That work would produce a Draft Constitution 0.1 built from the open questions, failure conditions, historical lessons, technical evidence, competing proposals, and practical expertise identified across the ecosystem.

The larger question is whether powers already distributed across governments, AI developers, infrastructure providers, evaluators, markets, and the public can be coordinated strongly enough to prevent catastrophic action, while remaining divided and constrained so that no government, company, coalition, or governing body controls the full decision chain.

Participation can begin with one question, source, correction, objection, or connection. It can grow into a project, working group, organizational profile, formal response, collaborative research effort, or contribution to the constitutional design process.

The alliance is intended to help people understand what already exists, recognize opportunities to combine strengths, and act on important problems that might otherwise remain between institutions.