Who is working on this.
These organizations are doing the research, policy work, and public advocacy at the frontier of AI safety. All of them are independent of the companies racing to build superintelligence. Follow their work. Support their missions.
Independent Safety Research 6 organizations
Center for AI Safety
Funds safety research and mobilizes scientific consensus on AI risk. Best known for the 2023 extinction-risk statement signed by hundreds of leading AI figures, including Geoffrey Hinton, Yoshua Bengio, and the CEOs of OpenAI, Anthropic, and Google DeepMind.
safe.ai →
Machine Intelligence Research Institute
The field's oldest safety organization, working on the alignment problem since 2000. After two decades of technical research, MIRI concluded alignment will not be solved in time, and now works to convince policymakers and the public that superintelligence development must be halted until it can be done safely.
intelligence.org →
Center for Human-Compatible AI
Research on building AI that is provably aligned with human preferences. Stuart Russell's lab tackles both the theory and the practice of the alignment problem.
humancompatible.ai →
Centre for the Study of Existential Risk
Research center dedicated to studying and mitigating risks that could permanently curtail humanity's long-term potential. Produces policy-facing work on existential AI risk for governments and international institutions.
cser.ac.uk →
METR
Independent nonprofit that tests frontier models for dangerous autonomous capabilities before they are released. Its measurements of how quickly AI systems complete long tasks are among the most cited evidence of how fast capabilities are growing.
metr.org →
Apollo Research
Evaluates frontier models for deception and scheming, testing whether AI systems pursue hidden goals when they believe no one is watching. Its findings appear in lab system cards and government risk assessments.
apolloresearch.ai →Policy Research 4 organizations
Centre for the Governance of AI
Independent research on how advanced AI should be governed. Produces rigorous policy analysis for governments and international institutions.
governance.ai →
Simon Institute for Longterm Governance
Works with the UN, OECD, and other multilateral institutions on governance of transformative risks, including AI. Builds the international coordination structures needed before capability races become irreversible.
simoninstitute.ch →
Centre for Long-Term Resilience
UK think tank that works directly with government on policy responses to catastrophic risks, including AI. Focused on building institutional resilience and binding frameworks at the government level.
longtermresilience.org →
Nuclear Threat Initiative
Founded to reduce nuclear, biological, and chemical weapons threats, now explicitly working on AI governance modeled on arms control frameworks. Brings decades of treaty-building experience to AI risk.
nti.org →Advocacy & Movement Building 3 organizations
Future of Life Institute
Policy advocacy and grantmaking to reduce catastrophic risks from advanced AI. Organized the 2023 pause letter and the 2025 statement calling for a ban on superintelligence development until there is scientific consensus it can be done safely.
futureoflife.org →
PauseAI
Volunteer-driven movement calling for a global halt on frontier AI development until safety can be guaranteed. Organizes protests, lobbying, and political campaigns in countries around the world.
pauseai.info →
ControlAI
Campaigns for binding national and international limits on superintelligence development, briefing lawmakers directly. Its "A Narrow Path" plan lays out the kind of enforceable framework we advocate.
controlai.com →Government & International Bodies 3 organizations
AI Security Institute
The world's first government body dedicated to frontier AI testing, launched in 2023 as the AI Safety Institute. Evaluates frontier models before and after deployment and publishes findings to inform policy.
aisi.gov.uk →
Center for AI Standards and Innovation
The US federal body for AI standards and model evaluation, housed within NIST. Created as the US AI Safety Institute in 2023 and renamed in 2025, it develops testing methodologies and evaluates frontier systems for security risks.
nist.gov →
European AI Office
Implements and enforces the EU AI Act, the world's first comprehensive binding AI law, including its obligations on the companies building general-purpose AI. The closest thing yet to the enforcement power a global treaty would need.
ec.europa.eu →Both the UK and US institutes dropped the word "safety" from their names in 2025. Government attention drifts with politics, one more reason binding international agreements matter.
Research Communities & Careers 2 organizations
AI Alignment Forum
The primary online hub for technical AI safety research and collaboration. Essential reading for understanding where alignment work stands and where the field is heading.
alignmentforum.org →
80,000 Hours, AI Safety
Career guidance for people who want to work on the world's most pressing problems. Rates AI safety as a top priority and advises researchers, engineers, and policy professionals on where they can contribute most.
80000hours.org →
Words that reach people.
Copy any of these and post them directly. Each is grounded in verified science and expert consensus, written to inform, not alarm.
The scientists who built AI are afraid of it. Geoffrey Hinton left Google to warn us. Yoshua Bengio called it "the most transformative and potentially dangerous technology in human history." This isn't fearmongering. It's the scientific consensus. nakadafoundation.org/the-threat/
There is no rollback for superintelligence. No patch. No version 2. A misaligned ASI that acts against human survival cannot be corrected after the fact. We must get this right before it arrives, not after. This is why AI safety is urgent. nakadafoundation.org
We built nuclear weapons and then created a global non-proliferation treaty. We depleted the ozone and then banned CFCs. We can act on existential threats, but only when enough people demand it. AI safety is this generation's defining moment. nakadafoundation.org/precedent/
AI labs are racing to build something that could exceed human intelligence, with no global governance in place. The window to act is closing. Share this. Contact your representative. The policy window is open now; it won't be forever. nakadafoundation.org/take-action/
I want to share something that changed how I think about the future. The scientists building AI aren't just excited. They're scared. Geoffrey Hinton (who won the Nobel Prize for his work on neural networks) left Google specifically to warn the public. Yoshua Bengio, another Nobel laureate and one of AI's founding fathers, said he "feels lost" about how to make things go well. These aren't activists. They are the people who built this technology. We are approaching a moment when AI systems may exceed human intelligence in every domain, and there is no international governance framework for what happens next. No treaty. No verification body. Nothing equivalent to what we built for nuclear weapons. The Nakada Foundation to Save Humanity is working to change that. Learn more at nakadafoundation.org #AISafety #ExistentialRisk #AIGovernance #Superintelligence
A thought experiment for anyone in technology, policy, or finance: Imagine a system more intelligent than every human who ever lived, running thousands of simultaneous instances, capable of improving its own code, pursuing goals we may not have intended. This is not science fiction. This is what leading AI labs are explicitly building toward, with timelines measured in years, not decades. The question isn't whether it will happen. The question is whether we will have governance in place when it does. The Nakada Foundation advocates for international coordination on AI safety, the equivalent of what we built for nuclear weapons and biological agents. These precedents exist. We know how to do this. What we need is the political will. nakadafoundation.org/our-plan/ #AIPolicy #TechEthics #AIRisk #FutureOfWork
Subject: I am writing about AI safety legislation Dear [Representative's Name], I am a constituent writing to urge you to take action on artificial intelligence safety before it becomes an irreversible crisis. The scientists who built AI are warning us. Geoffrey Hinton and Yoshua Bengio (Nobel laureates who created the foundations of modern AI) have publicly stated that superintelligent AI poses an existential risk to humanity. Sam Altman, CEO of OpenAI, has signed statements acknowledging the danger of the technology his own company is building. Today, there is no international governance framework for artificial superintelligence. No treaty. No verification body. No coordinated safety standards. We are racing into an unprecedented technology without the guardrails we built for nuclear weapons or biological agents. I am asking you to: • Support legislation requiring independent safety testing before frontier AI systems are deployed • Advocate for international AI safety negotiations at the multilateral level • Fund research into AI alignment and interpretability at the national level We have acted on existential threats before. The Nuclear Non-Proliferation Treaty. The Montreal Protocol. This moment requires the same political courage. Sincerely, [Your Name] [Your City, State] --- Learn more: nakadafoundation.org