A field guide to meaningful AI work
Want to help with AI safety, but not sure how?
Explore paid roles, volunteer projects, fellowships, and emerging initiatives across the organizations shaping AI safety. Get matched to work that fits your skills—or register your interest in a future community across labs and countries.
PROJECTS & OPPORTUNITIES LIVE
People are looking for your help.
Current roles and projects, pulled from official career sites.
01
Research Engineer, InterpretabilityApollo Research · London · Hybrid
Open02
Autonomy Evaluations EngineerMETR · Berkeley · On-site
Open03
AI Safety Course FacilitatorBlueDot Impact · Remote
OpenMEET THE FIELD
See who's working on what.
Independent evaluators, research labs, public institutes, standards bodies, and training programs.
Advances AI safety and governance through research, policy advice, risk monitoring, and international dialogue, with deep knowledge of China’s AI ecosystem.
Measures frontier-model capabilities, autonomy, and loss-of-control risks.
Develops AI control methods, studies strategic deception, and investigates risks from misaligned AI agents.
Studies how to navigate rapid AI progress and the transition to a world with superintelligent systems.
Researches scheming, deception, evaluations, and frontier AI security.
Incubates and conducts technical AI safety research and red-teaming.
Conducts safety research, builds the field, and develops safety standards.
Funds and connects research on cooperation, coordination, and multi-agent safety.
Studies how advanced AI systems can reliably act in line with human values.
Runs open research programs that help new contributors produce useful safety work.
Offers AI safety and governance courses, community, and career support.
Funds ambitious R&D, including large programs in scalable AI trust and coordination.
MISSION-DRIVEN DOESN'T MEAN UNDERPAID
Do important work—and get paid very well.
Technical safety, security, and evaluation teams compete for exceptional talent. A move into AI safety can be a serious career step, not a financial sacrifice.
Explore live compensation & role stats →FRESH FROM OFFICIAL CAREER SITES
Recent jobs and opportunities
6 current roles · crawled every six hours
A WORKING MAP OF THE FIELD
Explore the challenges shaping AI safety.
We're documenting the field problem by problem—what each challenge is, the questions that remain open, and who is already working on it.
Browse all 8 challenges →01
Evaluating advanced AI
Build credible tests for dangerous capabilities, strategic behavior, autonomy, and safeguards before systems are widely deployed.
4 organizations mappedExplore →
02
Understanding how models work
Develop methods that reveal what models represent, how they reason, and when their internal processes differ from their stated behavior.
4 organizations mappedExplore →
03
Controlling autonomous systems
Keep increasingly capable agents within legitimate authority, even when they operate for long periods, use tools, or coordinate with other agents.
4 organizations mappedExplore →
04
AI security and misuse
Prevent model theft, malicious use, compromised agents, and cascading failures across the infrastructure on which advanced AI depends.
4 organizations mappedExplore →
YONDON'S ECOSYSTEM MAP
AI safety projects
Research source updated August 17, 2026
See who is doing what, why they matter, and the highest-leverage way to help.
Explore projects & organizations →ECRI AI error reporting
AI incident reporting and independent patient-safety investigation
View initiative →AI Agency Protocol
Purpose-bound, expiring authority for autonomous AI agents
View initiative →Frontier AI Critical Defense Program
Virtual patching for AI-discovered critical-infrastructure vulnerabilities
View initiative →Democratic AI oversight initiative
Capacity, training, and tools for oversight of government AI use
View initiative →NO CAREER PIVOT REQUIRED
You can get involved before you change jobs.
Start with a problem you genuinely care about, people you want to learn from, or a piece of work that needs doing. There is no prescribed sequence.
01
UNDERSTANDStart with the problems
8 challenges →02See the questions teams are trying to answer—and who is already working on them.
MEETFind people around one challenge
Community →03Help shape future conversations across labs and countries around shared AI safety problems.
DOPick up something concrete
6 openings →Browse paid roles, fellowships, volunteer work, and scoped projects from active organizations.