Map of AI Existential Safety

Summary

This seeded note records the AISafety.com map as a directory-style raw source rather than a technical paper. The clip contains an SVG map and more than 300 entries for organizations, projects, programs, media channels, funders, training programs, and other AI existential-safety resources. The source is useful as field-navigation context, not as direct evidence for a technical claim.

Key Claims

• The page presents itself as an overview of key organizations, programs, and projects operating in the AI safety space.
• Its category vocabulary includes training and education, funding, governance, empirical research, research support, blogs, resources, conceptual research, advocacy, strategy, forecasting, media, career support, and capabilities research.
• The map treats AI safety as a field ecosystem rather than a single research agenda, spanning technical research, governance, advocacy, funding, communication, and career pathways.
• Because the source is a live directory, individual entries and organizational descriptions may change and should be rechecked before use in current recommendations.

Methods / Formalism

• This is not a formal model or empirical benchmark.
• The useful structure is a taxonomy over field roles. In the clipped May 2026 version, the most common single categories are Training and education, Funding, Governance, Empirical research, Research support, Blog, Resource, Conceptual research, Advocacy, Video, Strategy, and Career support.
• Treat the map as a routing layer: it can help locate institutions or programs relevant to Strategic Reasoning, AI Evaluation and Benchmarking, Safe Multi-Agent Reinforcement Learning, or Safe Reinforcement Learning, but it does not by itself settle technical questions in those areas.

Evidence / Experiments

• The source is an edited public directory and visual map.
• The current ingest reviewed the page title, summary text, headings, and category distribution, not every listed organization one by one.
• Organization descriptions should be treated as directory metadata and verified against primary sources before being used in high-stakes or current claims.

Connections

• Connects to Strategic Reasoning through governance, strategy, advocacy, and field-building resources.
• Connects to AI Evaluation and Benchmarking through organizations and projects working on empirical safety, evaluations, and capability assessment.
• Connects to Safe Multi-Agent Reinforcement Learning and Safe Reinforcement Learning at the field-navigation level, especially for locating technical safety programs and support infrastructure.
• The source is intentionally not expanded into a concept page because its main payload is a changing directory, not a stable technical object.

Open Questions

• Which organizations in the directory have primary technical papers or benchmark releases that should be ingested separately?
• Should the wiki eventually maintain an output page for field-navigation resources, distinct from durable concepts and sources?
• Which parts of the directory are stable enough to cite, and which require live web verification?

Citation

AISafety.com. (2026). Map of AI Existential Safety. AISafety.com.