01
Navigate the field
Make sense of the landscape with curated maps, small-group conversations, and people who can answer the questions search cannot.
A COMMUNITY FOR FINDING YOUR WAY IN
Navigate the field.
Coordinate what matters.
AI safety is a collection of hard problems, organizations, and people who often need better ways to find one another. We’re forming a practical community to help people learn the landscape and work together across the field’s major challenges.
Cross-disciplinaryChallenge-ledProject-oriented
THE COMMUNITY, IN PRACTICE Forming now
Not another feed to keep up with.
A place to develop context, relationships, and shared work—with enough structure to help people move from interest to contribution.
01
Field navigationOrientation, maps, and practitioner context
02
Challenge circlesSmall groups around consequential problems
03
Useful introductionsPeople, projects, organizations, and opportunities
04
Coordination supportTurn discussions into scoped next steps
A COMMUNITY WITH A JOB TO DO
Help people become more useful to the field—and to each other.
Good community infrastructure should reduce the cost of orientation and coordination. These are the three outcomes we would optimize for first.
02
Find your collaborators
Meet people across research, engineering, policy, operations, and field-building—especially those working on the same problem from another angle.
03
Coordinate useful work
Turn shared concern into reading groups, scoped projects, introductions, events, and contributions that existing organizations can actually use.
COORDINATE AROUND THE PROBLEMS
Start with a challenge you care about.
Challenge circles give the community a shared object of attention. They can connect newcomers, active practitioners, and project organizers without forcing everyone into the same career path.
01Capability evaluations
Evaluating advanced AI
Build credible tests for dangerous capabilities, strategic behavior, autonomy, and safeguards before systems are widely deployed.
QUESTION TO COORDINATE AROUNDHow can we tell when an AI system has crossed a meaningful risk threshold?
02Mechanistic interpretability
Understanding how models work
Develop methods that reveal what models represent, how they reason, and when their internal processes differ from their stated behavior.
QUESTION TO COORDINATE AROUNDCan we inspect advanced systems well enough to catch deception or dangerous reasoning?
03Control protocols
Controlling autonomous systems
Keep increasingly capable agents within legitimate authority, even when they operate for long periods, use tools, or coordinate with other agents.
QUESTION TO COORDINATE AROUNDHow do we preserve meaningful human control as agents become more capable and independent?
04Cyber evaluations
AI security and misuse
Prevent model theft, malicious use, compromised agents, and cascading failures across the infrastructure on which advanced AI depends.
QUESTION TO COORDINATE AROUNDWhich defenses still work when both attackers and defenders can use capable AI agents?
05AI-assisted oversight
Scalable oversight
Help people reliably evaluate work that is too fast, complex, or specialized for unaided human reviewers to check directly.
QUESTION TO COORDINATE AROUNDHow can humans supervise systems whose outputs exceed our own ability to verify them?
06Risk standards
Governance and safety standards
Turn evidence about AI risk into enforceable rules, measurable standards, institutional capacity, and accountable deployment decisions.
QUESTION TO COORDINATE AROUNDWhich rules and institutions can keep pace with rapidly advancing capabilities?
07Cooperation research
Multi-agent coordination
Understand cooperation, conflict, bargaining, and emergent behavior when many AI agents and human institutions interact.
QUESTION TO COORDINATE AROUNDHow can independently developed agents cooperate without collusion, conflict, or systemic failure?
08Joint evaluations
International coordination
Create shared testing practices, risk thresholds, and response capacity across countries without reducing safety to the weakest consensus.
QUESTION TO COORDINATE AROUNDWhat can governments coordinate on before the highest-risk systems are globally deployed?
THERE IS MORE THAN ONE WAY IN
Come with curiosity, experience, or something to organize.
01
I’m exploringLearn the vocabulary, compare directions, and meet practitioners without pretending you already know your path.
02
I’m already workingFind peers outside your organization, share useful context, and coordinate where the field has gaps.
03
I’m organizingBring a concrete question, project, event, or need—and find people who may be able to move it forward.
HELP SHAPE THE FOUNDING GROUP
Tell us what would make this community genuinely useful.
Share what concerns you, what you are working on, and the kinds of people or projects you hope to find. We’ll use that to form the first conversations and challenge circles.